Drexel Burnham Lambert, Underwriting Fees, and Market Power

Home , Drexel Burnham Lambert, Michael Milken

DREXEL BURNHAM LAMBERT, UNDERWRITING FEES, AND MARKET POWER

GLENN R. WILLIAMS

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2004

Glenn R. Williams

This document is dedicated to my parents, Horace and Mildred Williams, and my sons, Benjamin, Daniel and Samuel Williams.

ACKNOWLEDGMENTS

Miles Livingston provided endless encouragement and direction in this endeavor.

David Brown, Joel Houston, and Jay Ritter also provided helpful comments and suggestions. My fellow graduate students, especially Stas Nikolova and Rick Borghesi, provided invaluable insight and support. Seminar participants at Florida Atlantic

University and the University of Central Florida also added helpful comments.

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... iv

LIST OF TABLES...... vi

LIST OF FIGURES ...... vii

ABSTRACT...... viii

CHAPTER

1 BACKGROUND MATERIAL ...... 1

Literature Review and Development of Testable Hypothesis ...... 3 Market Concentration...... 4 Underwriting and Selling securities ...... 6 Junk Bonds ...... 8 A Summary of Our Findings from Previous Research...... 9 Testable Hypothesis...... 9 History of the Junk Market: 1977-1990...... 11 Junk Bonds as a Different Market from Investment Grade Issues ...... 13 Data...... 14 Summary Statistics ...... 15

2 UNDERWRITING FEES...... 24

Theoretical Model...... 24 Empirical Analysis...... 31 Regression Analysis...... 35 Underwriting Fees ...... 35 Treasury Spreads ...... 39

3 COMMERCIAL BANK ENTRY...... 53

4 CONCLUSION...... 63

LIST OF REFERENCES...... 65

BIOGRAPHICAL SKETCH ...... 68

LIST OF TABLES

Table page

1-1 Summary statistics for junk issues ...... 18

1-2 Summary statistics for B-rated issues...... 19

1-3 Summary statistics for investment-grade issues...... 20

1-4 Summary statistics for A-rated issues ...... 21

1-5 OLS regression of underwriting spreads of junk bonds 1977-1999...... 22

1-6 OLS regression of underwriting spreads of investment-grade bonds ...... 23

2-1 Market share of investment-grade underwriters 1977-1999...... 41

2-2 Market share of junk underwriters 1977-1999...... 42

2-3 Summary statistics for the junk bond market from 1977-1989...... 43

2-4 Percent of total junk principal issued by the top 10 underwriters...... 44

2-5 OLS regression of underwriting spreads for junk bonds...... 45

2-6 OLS regression to determine difference between underwriter spread of the top underwriter and those of all other underwriters 1977-1989...... 46

2-7 OLS regression of treasury spreads for junk bonds ...... 48

2-8 OLS regression to determine difference between treasury spreads of the top underwriter and those of all other underwriters 1977-1989...... 49

3-1 Summary statistics: Investment vs. commercial banks...... 58

3-2 Commercial bank underwriters 1991-1999...... 59

3-3 Percent share of the major commercial bank junk bond underwriters 1991-1999...60

3-4 Regressions with commercial bank (CB) dummy for issues 1991-1999...... 61

3-5 Regressions with commercial bank market share 1991-1999...... 62

LIST OF FIGURES

Figure page

2-1 Median underwriter fees as a percent of principal issued...... 51

2-2 Median junk bond underwriter fees as a percent of principal 1977-1989...... 52

vii

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

DREXEL BURNHAM LAMBERT, UNDERWRITING FEES, AND MARKET POWER

Glenn R. Williams

August 2004

Chair: Dr. Miles Livingston Major Department: Finance

I examine the market for new issue bonds from 1977-1999 and show that underwriting fees in the 1980s junk bond market were higher due to the dominant position of Drexel Burnham Lambert. The market for underwriting junk was much more competitive following Drexel’s bankruptcy in 1990 and underwriting fees were lower. In contrast to other researchers, I do not find evidence of competitive pressures on underwriting fees from commercial banks entering the new issue junk bond market.

viii CHAPTER 1 BACKGROUND MATERIAL

Since 1977 we have seen the junk bond market reborn, grow to a multi-billion dollar business, endure scandal and the loss of its dominant underwriter, and finally mature into an accepted method for raising capital. More than a decade has passed since the junk bond market scandals and Drexel Burnham Lambert’s filing for bankruptcy in

1989--sufficient time to examine the legacy of the junk bond market and to make judgments about the workings of Michael Milken, Drexel Burnham Lambert and the market for newly issued junk bonds.

There is a wealth of anecdotal evidence stating that the 1980s junk bond market was different (see Stewart (1992), Bruck (1989), and Stein (1992)). Most of this evidence points to the dominance of Michael Milken in creating and then controlling the market.

Bruck (1989, page 57) reports the observation of a junk buyer after meeting with Milken in 1979:

He had the issuers. He had the buyers. He had the most trading capital of any firm. He had the knowhow. He had the best incentive system for his people. He had the history of data--he knew the companies, he knew their trading prices, probably their trading prices going back at least to 1971. He had boxed the compass.

The accusation has also been made that Milken had “captive clients.” That is, that he had a set of buyers ready and willing to buy whatever junk issues he made available to them. As Stewart (1992, page 54) reports:

By early 1977, Milken’s operation controlled a remarkable 25% of the market in high-yield securities. It was really the only firm maintaining an active market- making operation with an eye toward enhancing the liquidity of the market.… So Milken became, in effect, the market for high-yield bonds. He had an incredible

1 2

memory, and he knew who owned what issues, what they had paid, their yield to maturity, and who else wanted them. Increasingly, his clients developed such confidence in his research and market acumen that when he urged them to invest in a particular issue, they did.

Almost 30 years after issuing his first junk bond, Michael Milken continues to inspire heated debate and diverging opinions. As the Wall Street Journal editorialized:

No doubt historians someday will set the record straight on Mr. Milken's contribution to the explosive economic growth experienced by the U.S. in the past 20 years. For now, he remains the most famous figure who became engulfed in the controversial securities prosecutions undertaken by then-U.S. attorney Rudolph Giuliani, based on information supplied by such economically immaterial figures as Ivan Boesky, Dennis Levine and Martin Siegel. (“Yes, Pardon Milken,” Wall Street Journal, December 13, 2000, page A26.)

This paper is a beginning in setting the record straight, at least in part, by examining the effect Drexel had on the competitive structure of the 1980s junk market and how the market changed in the 1990s. The market for new issue junk is well suited for a study of the relationship between market concentration and competitive behavior because of the exogenous shock in market concentration that occurred when Drexel

Burnham Lambert liquidated in 1990. Drexel re-created the new issue junk market in the late 1970s, dominated it for a decade, and then made a sudden, unexpected exit in 1990.

This provides the advantage of comparing competition in the same market across time without the confounding influences of other factors that would arise if Drexel’s market share had eroded over several years. We examine the effect of this shock on market concentration and underwriting fees.

Our results add to the literature documenting noncompetitive behavior among banks in concentrated markets (see Berger, Demsetz and Strahan (1999) for a summary of this literature). We also add to a growing list of work examining underwriting fees (see

Kim, Palia and Saunders (2003), Chen and Ritter (2000), Livingston, Pratt and Mann

(1995)). We continue this strand of literature by offering explanations for the observed

decline in underwriting fees for debt issues in the early 1990s and examining how the

junk market changed following Drexel’s exit.

We see public policy implications in our work as well. Previous research by Gande,

Puri and Saunders (1999) suggested that allowing commercial banks into the market for underwriting corporate debt in 1991 resulted in lower underwriting fees. Our results suggest that the role of commercial banks in lowering fees at that time is overstated. We test the hypothesis that improved competition from more firms in general and not simply a different type of firm (in this case, commercial instead of investment banks) is the reason corporate bond underwriting fees are lower in 1991.

We find that Drexel exerted significant market power over the junk market in the

1980s and used this power to charge high underwriting fees. We attribute this to Drexel’s ability to signal higher quality through their large position in the aftermarket and their pioneering role in new issue junk. Other underwriters began to match Drexel’s fees in the late 1980s and there is evidence of some fee clustering for junk bond issues late in the decade around 3.50 percent. Following Drexel’s liquidation in 1990, the leading underwriters could not make the same quality signals as Drexel and were unable to maintain higher fees. Unlike previous researchers, we do not attribute lower fees to commercial bank entry. We find that commercial bank fees are no different than investment bank fees, and the market share of commercial banks is too small prior to

1996 to account for the dramatic drop in fees in the early 1990s.

Literature Review and Development of Testable Hypothesis

To develop our testable hypothesis, we draw upon several strands of literature concerning market structure, competition and innovation. The literature relating to our

paper falls into one of three broad categories: market concentration, junk bonds, and the

market for underwriting and selling securities. Among the market concentration literature, we focus on the large amount of research concerning market concentration among banks, particularly as it relates to the interest paid on demand deposits. The work on junk bonds focuses on how junk differs from investment grade bonds and previous studies of issues by Drexel Burnham Lambert. For the underwriting process, most of the work has investigated equity issues, but the results about the competitive nature of that process are easily adapted to our work on debt issues.

Market Concentration

We begin by considering several papers that investigate the relationship between market concentration and the pricing of deposit rates. Gilbert (1984) provides a summary of the early literature covering this topic, while Berger, Demsetz and Strahan (1999) provide a more recent summary article. This research is relevant to our study of the underwriting market because it provides empirical evidence for the relationship between market concentration and prices in financial markets and it discusses the theoretical underpinnings of market concentration and pricing.

Two competing hypotheses are found in the economic literature investigating the relationship between firm profits and market concentration. First, the structure- performance hypothesis states that firms in concentrated markets maximize profits through anti-competitive actions to keep prices high. In this framework, high prices and profits are the result, and not the cause, of market concentration. Note that for this structure to hold there must be some barriers that keep potential competitors out; otherwise the high level of profits will attract new entrants.

The counter-argument to the structure-performance hypothesis is the efficient-

structure hypothesis discussed by Demsetz (1973). He points out that market

concentration may sometimes be the result of competitive behavior from more efficient

firms. For example, consider 2 firms, X and Y, who are making a uniform product. If

firm X has lower marginal costs than Y, then X will maximize profits by charging a price

lower than Y. In this scenario, the higher market share of X is the result of competitive

actions that are beneficial to consumers.

The efficient-structure hypothesis, however, need not always result in lower prices

for the consumer. Consider the same firms X and Y, but now suppose they produce a

good that has substantial quality differences. In this case, if X has an efficiency advantage by delivering higher quality at the same cost, it may capture market share while charging a price equal to, or even higher than, firm Y.

In summary, both the structure-performance hypothesis and the efficient-structure hypothesis predict that the leaders in market share will enjoy higher profits. While the structure-performance hypothesis predicts higher prices from firms with the highest market share, relative prices in the efficient-structure scenario depend on the quality differences in competing products.

As a test of these competing theories of market concentration, banking research has focused on deposit rates. Deposit rates in the 1980s provided an ideal laboratory to study market concentration and competition because of laws that limited intra-state banking. As

Berger and Hannan (1989, page 291) point out, “… the existence of geographically limited markets offers researchers the advantage of comparing profitability and concentration across markets without the confounding influences of interindustry

6 differences.” Since the rate paid to depositors is a homogenous product that cannot be differentiated by quality, the two theories result in different predictions for prices. If the structure-performance hypothesis holds, one would expect to find higher prices (lower deposit rates) in more concentrated markets. In an efficient-structure framework, however, highly concentrated markets are the result of lower prices (higher rates) from the most efficient banks.

Using a sample of deposit rates from 1983-1985, Berger and Hannan (1989) find evidence to support the structure performance hypothesis. They report interest rates paid to depositors in the most concentrated markets were 25 to 100 basis points less than those paid in the least concentrated markets, suggesting that banks have strong market power over depositors. Hannan (1991) finds a similar result in commercial loan rates. Neumark and Sharpe (1992) add to this evidence of noncompetitive pricing by reporting that banks in concentrated markets are slower to raise their deposit interest rates in response to a rise in interest rates. This line of research supports the structure-performance hypothesis for the pricing of loans and deposit rates by banks in concentrated markets.

Underwriting and Selling Securities

Support for the efficient-performance hypothesis among investment banks is provided by Tufano (1989). He studied 58 financial innovations created between 1974 and 1986. Tufano points out that these innovations are expensive to develop, cannot be patented, and are easily imitated by competing investment banks. Yet he finds no evidence that innovating banks profit from higher fees to issue their products. Instead, he finds that innovators actually charge lower fees than imitators, even prior to imitators entering the market. Market share, however, is maintained by the pioneer, with the resulting gain in revenues. Tufano (1989, page 234) reports that

Syndicates led by pioneers earn average underwriting revenues 2.8 times as large as syndicates led by any other bank in the first year of their new products’ lives. Although innovators charge lower prices, they underwrite more deals and capture larger underwriting revenue than imitators.

Based on discussions with investment bankers, Tufano proposes that innovating investment banks gain market share in two ways. First, they use information gained on the early deals to gain a larger share of the secondary market in their new products.

Secondly, they signal higher quality by being the innovating bank in a security.

Evidence of noncompetitive behavior in securities markets is provided by Christie

and Schultz (1994). They study bid and ask quotes for 100 of the most active stocks in

the NASDAQ market in 1991 and find that odd-eighth quotes are virtually nonexistent.

They find no evidence that this is due to the negotiation hypothesis of Harris (1991) or to

trading volume or other factors that may influence spreads. Interestingly, in a follow-up

article Christie, Harris and Schultz (1994) report that NASDAQ dealers suddenly stopped avoiding odd-eighth quotes following publicity of their first article. The results support the claim of implicit collusion among NASDAQ dealers to profit from higher spreads.

Dutta and Madhavan (1997) follow this work by presenting a game-theoretic model showing that dealers may behave in a non-cooperative manner and set bid-ask spreads above competitive levels.

Lee, Lochhead, Ritter and Zhao (1996) study the costs of raising capital from 1990 to 1994. They find that underwriting spreads are highest for equity IPOs and lowest for straight debt IPOs, with seasoned equity and convertible debt falling in between. They also report substantial economies of scale, although not as much for debt offerings as equity issues. The credit quality of debt is shown to be a major factor for determining costs, with lower quality debt costing substantially more. The authors note the link

between issue size, credit quality and economies of scale: as the size of the issue

increases, all else equal, it will tend to drive down the credit rating and increase the fees,

masking some of the economies of scale.

Chen and Ritter (2000) find evidence of implicit collusion among investment

bankers in the equity IPO market. Their study examines equity IPOs from 1985 to 1998.

For the period from 1995-1998, they find that more than 90 percent of equity IPOs had

gross underwriting spreads of exactly 7 percent. The spreads are unrelated to the cost of

underwriting, since the spread on the riskiest IPOs are priced the same as easier to value

IPOs. Instead of the expected economies of scale, small IPOs spreads are equal to the

spread on large IPOs. They offer several explanations for the ability of underwriters to

charge higher than competitive spreads, including the importance of analyst coverage,

failure of the issuers to focus on costs, and the ease of justifying high spreads by referring

to previous deals done at 7 percent.

Junk Bonds

Livingston, Pratt and Mann study debt issues by Drexel Burnham Lambert. They find that Drexel charges higher underwriting fees than other underwriters, but that Drexel bonds have lower yield spreads at issue. The authors calculate the “cost to issue” as the yield to maturity of a bond at issue based on the proceeds from the issue after all fees are paid. On this basis, the cost of having Drexel underwrite the issue is the same as the cost from any other underwriter. Although Drexel gets a higher underwriting fee, they deliver more proceeds.

Jewell and Livingston (1998) study split rated bonds from 1980-1993 and find that yields for split rated bonds are generally the average of the two ratings. Underwriting fees for junk bonds are found to vary substantially by rating, while fees for all investment

grade bonds show little variation across ratings. Neither agency is found to give

consistently lower ratings, and they do not find evidence that investors value one rating

agency over the other.

Other studies have found that junk bonds have a substantial equity component.

Datta, Iskandar-Datta, and Patel (1997) find that junk bond IPOs, like equity IPOs, are underpriced, while investment grade debt IPOs are overpriced. Blume, Keim, and Patel

(1991) find that junk bonds have a low correlation with the returns on government securities and a higher correlation with common stocks. Cornell and Green (1991) find that changes in stock prices explain the variance of junk bond returns better than changes in interest rates.

A Summary of Our Findings from Previous Research

1. Rates on deposits and commercial loans in the 1980s provide evidence for the structure-performance hypothesis and non-competitive pricing among banks in concentrated markets.

2. Innovation among investment banks supports the efficient structure hypothesis (Tufano (1989)).

3. The preponderance of gross spreads of exactly 7 percent for equity IPOs from 1995-1998 is evidence of noncompetitive behavior among investment banks (Chen and Ritter (2000)).

4. Junk bonds differ from investment grade bonds in that they are more sensitive to ratings (Jewell and Livingston (1998)) and exhibit a substantial equity component (Fridson (1994)).

Testable Hypothesis

From a theoretical viewpoint, we draw upon the work of Berger and Hannan (1989) and Demsetz (1973) and assert that the concentrated junk market from 1977-1989 should follow either the structure-performance hypothesis or the efficient structure hypothesis.

The structure-performance hypothesis would predict that because of the highly

concentrated market, the more efficient firms were able to charge higher prices and glean

higher profits. On the other hand, the efficient-structure hypothesis would predict that the

most efficient firms cause higher concentration in the market by capturing a higher

market share through competitive actions.

Based on the anecdotal evidence, our hypothesis is that the market for underwriting junk from 1977-1989 is a special case of the efficient-structure hypothesis, where the most efficient firm gains market share not by capitalizing on its profitability at lower fees, but by using its quality advantage. Berger and Hannan (1989) point out this special case of the efficient structure hypothesis and note that it does not apply to deposit rates since they do not exhibit differences in quality. According to this hypothesis, Milken was able to signal quality to junk bond issuers and buyers through his position as the innovator and largest dealer in the market, and his history of providing investors superior returns from investments in fallen angels.

If Drexel’s dominance of the 1980s junk market is a special case of the efficient- structure hypothesis, we expect Drexel to extract rents for providing the highest quality service. We test this by comparing Drexel’s underwriting fees to the fees of all other underwriters in the 1980s junk bond market. We expect the ability to extract rents and higher prices to diminish after Milken and Drexel are out of the market. We test this by comparing the fees of the leading underwriter in the 1990s junk bond market to the fees of all other underwriters.

How other underwriters respond to Drexel’s quality advantage is not clear based on the theoretical work to date. On the one hand, if the structure-performance hypothesis holds, we would expect competing underwriters to have difficulty entering the market for

junk. If the structure-performance hypothesis is true, we would expect to find

underwriters able to enter the market at lower fees, unless they can signal quality as

Drexel does, at which point they can charge higher fees. We document the number and

pricing of Drexel’s competition from 1977-1989, and test whether Drexel’s fees were

different than other underwriters’ fees through time.

History of the Junk Market: 1977-1990

While a student at Berekely, Michael Milken read "Corporate Bond Quality and

Investor Experience, 1900 to 1950" by W. Braddock Hickman. Hickman’s study pointed out that a diversified portfolio of non-investment grade bonds outperformed investment grade bonds from 1900-1950. Due to a high number of defaults in the 1930s, the market for new issue high yield bonds was virtually nonexistent between 1940 and 1970. Most high yield bonds prior to 1977 were “fallen angels”--bonds that were issued as investment grade but later downgraded to junk status.

Milken continued his education at Wharton and interned at Drexel Firestone, where he stayed as a full-time employee after completing his MBA. Based on his examination of Hickman’s study, Milken realized fallen angels were selling at too large a discount and he profited handsomely by buying them. Over time, he convinced enough investors of the potential profits in junk bonds that he created a liquid market for reselling junk, and later a market for new issue junk bonds. He and his firm, Drexel Burnham Lambert, were highly rewarded for this innovation. In 1975 there were virtually no junk bonds issued, but in 1983 Drexel earned fees on $4.69 billion in new-issue junk bonds.

Drexel dominated the new issue junk market for a decade. In 1983, for example, when Drexel underwrote $4.69 billion in new issue junk, the next closest underwriter was

Salomon Brothers with $445 million. That year, Drexel was the underwriter for 50 out of

79 issues, and 66% of the total par value of bonds brought to market.

Such market dominance led to allegations of price fixing and exorbitant fees on the part of Drexel. It was also alleged that Milken and Drexel maintained “captive” clients for junk issues, buyers who were willing and able to purchase whatever Milken brought to the market, without the usual regard for the financial condition of the issuer.1 As

reported by Berman and Michaels (1992, page 78), Drexel and Milken countered

allegations of wrongdoing with claims of serving the capitalistic ideal:

“… Drexel used financial technology to create value, build businesses and jobs and consumer services by enabling entrepreneurs to raise money far more efficiently than they could have done without us.”

In the end, admissions of insider trading and other SEC rules violations led to the much-publicized demise of Drexel in the late 1980s. Milken pleaded guilty to securities fraud and spent time in prison. Drexel filed for bankruptcy in 1990, and the firm was voluntarily liquidated over the next several years.

In addition to the scandals at Drexel, the passage of the Financial Institutions

Reform Recovery and Enforcement Act (FIRREA) in 1989 required savings and loans to sell their junk bond holdings. That same year, an academic paper by Asquith, Mullins and

Wolff (1989) challenged the traditional view of junk bond default rates. Prior to their work, junk bond default rates were thought to be about 5%. Asquith et al. pointed out that the traditional way of calculating default rates, dividing the number of defaults in a year by the number of junk bonds outstanding, is not an accurate measure of default when the number of issues each year is increasing dramatically. They found that actual defaults

1 This accusation forms the basis of Stein’s A License to Steal.

rates were closer to 30% when considered over the lifetime of the bond. As evidenced by

Winkler (1990) their results were noted by practitioners and the popular press.

So in late 1989-early 1990 we have the collapse of Drexel, a re-evaluation of the default rate for junk bonds, and regulation requiring S&Ls to divest of their junk bond holdings. These three events combined to cause enormous downward pressure on prices and a sharp drop in the amount of new issue junk. In 1988 there were $31 billion in new non-convertible junk issues, but in 1990, the year of Drexel’s bankruptcy filing, new non- convertible issues of junk bonds had fallen to $1.4 billion. By October of 1990, the spread between junk bonds and 7 year Treasuries reached a record 12%, more than double the average spread in the 1980s.

This change of heart was short-lived, however, and by 1992 junk bonds were back up to almost $40 billion as other underwriters scrambled to fill the vacuum. In spite of the tainted legacy of the 1980s market, the financial community seems to have again embraced the idea of junk bond financing.

Junk Bonds as a Different Market from Investment Grade Issues

One reason we focus on the junk bond market is that the unique nature of the market has been overlooked in previous studies. Although previous studies note differences in underwriter fees due to credit quality, they proceed to make inferences about the entire bond market as a whole. This ignores the fact that the junk and investment grade bond markets are different--they serve different sets of investors with differing investment objectives.

Following the rebirth of junk in the late 1970s, the junk bond market has remained separate from investment grade bonds. One obvious reason is that the lower credit quality of junk bonds attracts different investors than the average investment grade investor. For

14 instance, many institutional investors, such as pension funds and insurance companies, classify junk bonds as “restricted securities” and are limited in the amount of junk they may own.

Previous research has found that junk bonds behave more like equity than investment-grade bonds. Datta, Iskandar-Datta, and Patel (1997) find that junk IPOs are underpriced, but investment-grade IPOs are overpriced. Blume, Keim, and Patel (1991) and Cornell and Green (1991) document that junk bond prices are more correlated with movements in stock prices than with interest rate changes.

Data

Our empirical sample includes all fixed coupon, non-convertible public debt issues from 1977–1999 for which data is available from the Securities Data Corporation database. We start the sample in 1977 because that is the year Drexel first issued a junk bond, and end the sample in 1999 to avoid any confounding influences in the underwriting market due to the repeal of Glass-Steagall in November 1999. Bonds rated

Baa or higher by Moody’s are classified as investment grade and bonds with a Moody’s rating of Ba or lower are classified as junk. If a bond is not rated by Moody’s, but has a

Standard and Poor’s (S&P) rating, it is assigned the corresponding Moody’s rating.

Bonds rated as investment by one agency and junk by the other are classified as junk. If the bond is not rated by either agency, it is not included in the sample. Following the previous literature (e.g. Gande et al. (1999)), I eliminate firms in regulated industries

(single digit SIC code 4) and financial firms (SIC Code 6). This results in a sample of

4,953 issues, with 3,437 investment grade bonds and 1,516 junk bonds.

Summary Statistics

Summary statistics showing the number of issues, average size, total size, and median underwriter fees for junk bonds are shown in Table 1-1; statistics for the investment-grade sample are shown in Table 1-3. Statistics are shown for all bonds, as well as the most frequent rating in each category (A for investment-grade issues (Table 1-

4) and B for junk issues (Table 1-2)). It is easy to see the tremendous growth in junk bonds from these statistics. From 1980 to 1989, the amount of junk principal in our sample increased over 17 times. By contrast, investment grade principal only almost doubled during this time. Also note the tremendous growth in our sample of junk in 1983: from $761 million in issues in 1982 to almost $2 billion in 1983. For the junk market as a whole in 1983, Bleakley (1984) reported that “… Wall Street underwrote more than $7 billion of the low-rated, high-yielding bonds, more than in the previous four years combined.”

Underwriter fees are markedly different between the two classes of bonds. Fees for underwriting junk are much higher, with yearly medians ranging from 1.8 percent of principal in 1990 to 3.5 percent of principal in 1987—1989. For investment grade issues, the median underwriter fee ranges from a low of 0.6 percent of principal in 1998 and

1999 to a high of 0.9 percent of principal from 1977-1979. To make a closer determination of this difference, we need to account for factors such as credit quality, time to maturity, and principal amount that have been shown to impact underwriting fees

(e.g., Tufano (1989) and Gande et al. (1999)). We run the following regression for our entire sample of junk bonds and also for investment grade bonds:

UnderwriterSpread = α 0 + β cr CreditDummies + β split SplitDummiess + β 2 Log(Maturity) +

β 3 Log(Size) + β 4 Callable + β 5 DebtIPO + β 6 Public + β ind IndustryDummies + (1)

β year YearDummies

The results are reported in Table 1-5 for junk bonds and Table 1-6 for investment- grade bonds. The results show that junk bond underwriter fees are much more sensitive to credit quality, while investment grades issues are more sensitive to time to maturity. All of the rating coefficients for junk are large and significant. Fees for investment issues show only a slight sensitivity in the lowest (Baa) rating, while the coefficients for higher ratings are small and insignificant. Time to maturity is not a significant factor in determining underwriting fees among junk issues, but for investment-grade bonds the coefficient on log(maturity) is relatively large (0.19) and significant. The regression results confirm our assertion that underwriting of junk bonds is priced differently than the underwriting of investment-grade issues.

The changes in the level of underwriting fees is also pertinent to our paper. Since the omitted year dummy is year99, all of the coefficients on the year dummies are relative to the fee level in 1999. For junk bonds, there is a dramatic decline in fees in 1990. The coefficient on year89 is 0.86, while on year90 the fee level is less than half of that (0.46).

Note that junk bond fees are uniformly much higher from 1984-1989, with coefficients on the year dummies ranging from 0.56 in 1985 to 0.86 in 1989. In the 1990s, fees for issuing junk are much lower, showing a gradual decline from a coefficient of 0.34 on the year90 dummy to 0.06 on the year98 dummy. A joint test of the coefficients on the year dummies for 1993-1997 indicates that they are not significantly different from each other

(F = 0.92, p-value = 0.45). In the junk bond market, we see a sudden drop in underwriting fees in 1990, and fees stay at fairly uniform levels from 1993-1997.

Investment-grade fees show the same decline from 1977-1999, but the differences are much less dramatic than the changes in the junk market. Coefficients on the year dummies in the 1980s range from 0.15 in 1984 to 0.00 in 1983, and then less than 3 basis points difference from 1999 levels is seen from 1991-1998. Fees do not change from

1989 to 1990, and the decline in 1991, while statistically significant at the 10% level,

(F=2.77, p-value=0.09) is a drop of only 5 basis points.

Table 1-1. Summary statistics for junk issues. Sample includes all fixed coupon, non- financial, non-regulated bonds issued 1977- 1999 and rated Ba or lower with yield data available from the SDC database.. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Ratings are assigned numerical values as follows: Baa = 4, Ba = 5, B = 6, Caa = 7. Split rated bonds are assigned the average of the two ratings. If the bond is not rated by either agency it is not included in the sample. Total Principal Issued and Mean Size are in millions of dollars. Mean Maturity is in years. Mean and median UW fee are the total underwriting fees for the issue as a percent of total principal issued. All Junk Issues Total Median Number of Principal Median Median Underwriter Median Year Issued Issues Issued Size Maturity Fee Rating 1977 11 298 25 15 2.5 5.5 1978 30 785 25 20 3.0 6.0 1979 23 690 25 20 3.3 6.0 1980 18 675 30 20 3.0 6.0 1981 12 770 63 20 2.6 6.0 1982 16 761 43 18 2.8 6.0 1983 25 1,875 60 15 2.6 6.0 1984 34 2,610 54 10 3.3 6.0 1985 63 5,848 50 10 3.0 6.0 1986 113 17,126 100 10 3.4 6.0 1987 80 14,861 125 10 3.5 6.0 1988 65 11,813 140 10 3.5 6.0 1989 55 11,885 169 10 3.5 6.0 1990 4 650 150 8 1.8 5.3 1991 29 7,034 150 10 2.0 5.0 1992 133 21,238 135 10 2.5 6.0 1993 170 29,085 125 10 2.8 6.0 1994 91 15,426 115 10 2.8 6.0 1995 62 11,909 135 10 2.5 5.5 1996 102 18,235 150 10 2.9 6.0 1997 137 25,920 150 10 2.8 6.0 1998 177 35,374 150 10 2.8 6.0 1999 66 15,829 188 10 2.5 6.0 1516 250,694

Table 1-2. Summary statistics for B-rated issues. Sample includes all fixed coupon, non- financial, non-regulated bonds issued 1977- 1999 rated B for which yield data is available from the SDC database. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. If the bond is not rated by either agency it is not included in the sample. Total Principal Issued and Mean Size are in millions of dollars. Mean Maturity is in years. Mean and median UW fee are the total underwriting fees for the issue as a percent of total principal issued.

B-Rated Issues Total Median Number of Principal Median Median Underwriter Year Issued Issues Issued Size Maturity Fee 1977 3 95 30 20 3.0 1978 23 583 20 20 3.3 1979 18 487 23 20 3.3 1980 14 475 28 20 3.0 1981 9 510 60 20 2.8 1982 12 491 30 13 2.8 1983 14 690 45 15 3.0 1984 20 1,075 50 15 3.4 1985 40 3,655 50 10 3.2 1986 66 9,301 90 10 3.4 1987 53 10,873 125 10 3.5 1988 48 9,315 150 10 3.5 1989 42 9,779 173 10 3.5 1990 2 350 175 9 3.0 1991 5 735 150 7 3.0 1992 78 12,738 126 10 2.8 1993 113 17,640 125 10 2.8 1994 60 10,052 108 10 2.8 1995 26 3,889 124 10 3.0 1996 61 10,532 150 10 3.0 1997 100 17,835 150 10 3.0 1998 105 16,592 125 10 2.9 1999 35 7,319 150 10 2.8 947 145,010

Table 1-3. Summary statistics for investment-grade issues. Sample includes all fixed coupon, non-financial, non-regulated bonds issued 1975-2002 and rated Baa or higher with yield data available from the SDC database. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Ratings are assigned numerical values as follows: Aaa = 1, Aa = 2, A = 3, Baa = 4. Split rated bonds are assigned the average of the two ratings. Total Principal Issued and Mean Size are in millions of dollars. Mean Maturity is in years. Mean and median UW fee are the total underwriting fees for the issue as a percent of total principal issued.

Investment Grade Issues Total Median Number of Principal Median Median Underwriter Median Year Issued Issues Issued Size Maturity Fee Rating 1977 35 3,801 100 25 0.9 3.0 1978 22 2,615 100 28 0.9 3.0 1979 33 5,285 125 25 0.9 2.5 1980 85 10,993 100 10 0.8 3.0 1981 62 11,555 175 15 0.7 3.0 1982 84 9,362 100 10 0.7 3.0 1983 49 5,070 100 20 0.8 3.0 1984 57 8,635 100 10 0.7 3.0 1985 128 18,089 100 10 0.7 3.0 1986 230 35,374 100 10 0.7 3.0 1987 147 20,725 100 10 0.7 3.0 1988 118 20,403 138 10 0.7 3.0 1989 108 19,189 150 10 0.7 3.0 1990 105 20,444 150 10 0.7 3.0 1991 201 38,723 175 10 0.7 3.0 1992 214 45,088 200 10 0.7 3.0 1993 234 46,294 150 10 0.7 3.0 1994 101 16,146 150 10 0.7 3.5 1995 218 35,045 150 10 0.7 3.3 1996 197 43,113 150 10 0.7 3.3 1997 311 49,041 130 10 0.7 3.0 1998 439 73,654 150 10 0.6 3.0 1999 259 71,371 200 10 0.6 3.5 3,437 610,014

Table 1-4. Summary statistics for A-rated issues. Sample includes all fixed coupon, non- financial, non-regulated bonds issued 1975-2002 and rated A with yield data available from the SDC database. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Total Principal Issued and Mean Size are in millions of dollars. Mean Maturity is in years. Mean and median UW fee are the total underwriting fees for the issue as a percent of total principal issued.

A-Rated Issues Total Median Number of Principal Median Median Underwriter Year Issued Issues Issued Size Maturity Fee 1977 9 775 75 25 0.9 1978 11 1,245 100 25 0.9 1979 12 1,450 100 25 0.9 1980 45 5,000 100 10 0.7 1981 36 5,950 150 17 0.7 1982 38 3,847 100 10 0.7 1983 15 1,801 100 25 0.7 1984 33 3,495 100 10 0.7 1985 64 8,518 100 10 0.7 1986 91 10,717 100 10 0.7 1987 55 7,858 100 10 0.7 1988 65 11,688 150 10 0.7 1989 44 7,150 125 10 0.7 1990 59 11,325 175 10 0.6 1991 115 20,838 150 10 0.7 1992 87 17,422 200 10 0.7 1993 83 15,294 150 10 0.7 1994 29 4,275 100 10 0.6 1995 78 12,903 125 10 0.7 1996 78 19,769 188 10 0.7 1997 159 23,383 100 10 0.6 1998 202 23,815 54 10 0.6 1999 97 24,232 200 10 0.7 1,505 242,748

Table 1-5. OLS regression of underwriting spreads of junk bonds 1977-1999. The sample consists of all bonds issued between January 1, 1977 and December 31, 1999 with yield data available from the SDC database rated as junk by Moody’s. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Bonds not rated by either agency are not included in the sample. The regression takes the following form: UnderwriterSpread = + β CreditDummies + β SplitDummiess + β Log(Maturity) + β Log(Size) α 0 cr split 2 3 + β 4 Callable + β 5 DebtIPO + β 6 Public + β ind IndustryDummies + β year YearDummies Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued. Credit Dummies are dummy variables for each Moody credit level. Splits are dummy variables indicating the bond was rated differently by each agency. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. DebtIPO is a dummy variable equal to 1 if the issue was the first debt issue by the firm since 1970, and equals 0 otherwise. Industry Dummies are dummy variables (not reported) for each one-digit SIC code. Year Dummies are dummy variables for each year from 1977-1999. *, **, *** indicate significance at the 10%, 5% and 1% level respectively. The omitted dummies are SplitBaa/Ba and year99

Junk Issues Parameter Estimate t-value Parameter Estimate t-value Intercept 0.88 5.55 *** year77 0.17 0.98 split1_2 year78 0.60 5.01 *** dummyAa year79 0.59 4.57 *** split2_3 year80 0.38 2.74 *** dummyA year81 0.13 0.80 split3_4 year82 0.07 0.46 split2_4 year83 0.35 2.87 *** dummyBaa year84 0.65 6.03 *** dummyBa 0.97 15.69 *** year85 0.56 6.13 *** split5_6 1.43 20.53 *** year86 0.71 8.99 *** dummyB 1.67 27.96 *** year87 0.84 9.98 *** split6_7 1.78 20.80 *** year88 0.82 9.27 *** dummyC 2.00 14.72 *** year89 0.86 9.39 *** lmat 0.05 1.15 year90 0.34 1.31 lsize -0.08 -4.01 *** year91 0.28 2.47 ** callable 0.26 5.78 *** year92 0.07 0.96 debt_ipo 0.16 5.70 *** year93 0.21 2.86 *** public -0.04 -1.44 year94 0.26 3.21 *** year95 0.30 3.37 *** year96 0.24 3.03 *** year97 0.17 2.26 ** year98 0.06 0.82 Obs: 1,515 R-Squared 0.65

Table 1-6. OLS regression of underwriting spreads of investment-grade bonds. The sample consists of all bonds issued between January 1, 1977 and December 31, 1999 with yield data available from the SDC database rated as investment- grade by Moody’s. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Bonds not rated by either agency are not included in the sample. The regression takes the following form: UnderwriterSpread = + β CreditDummies + β SplitDummiess + β Log(Maturity) + β Log(Size) α 0 cr split 2 3 + β 4 Callable + β 5 DebtIPO + β 6 Public + β ind IndustryDummies + β year YearDummies Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued. Credit Dummies are dummy variables for each Moody credit level. Splits are dummy variables indicating the bond was rated differently by each agency. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. DebtIPO is a dummy variable equal to 1 if the issue was the first debt issue by the firm since 1970, and equals 0 otherwise. Industry Dummies are dummy variables (not reported) for each one-digit SIC code. Year Dummies are dummy variables for each year from 1977-1999. *, **, *** indicate significance at the 10%, 5% and 1% level respectively. The omitted dummies are AAA and year99

Invesment Issues Parameter Estimate t-value Parameter Estimate t-value Intercept 0.20 5.43 *** year77 0.11 0.98 split1_2 0.05 1.18 year78 0.07 5.01 *** dummyAa 0.02 0.76 year79 0.07 4.57 *** split2_3 0.00 0.09 year80 0.06 2.74 *** dummyA 0.03 1.12 year81 -0.05 0.80 split3_4 0.04 1.59 year82 0.02 0.46 split2_4 0.07 0.61 year83 0.00 2.87 *** dummyBaa 0.08 3.09 *** year84 0.15 6.03 *** dummyBa year85 0.06 6.13 *** split5_6 year86 0.07 8.99 *** dummyB year870.089.98*** split6_7 year88 0.04 9.27 *** dummyC year890.059.39*** lmat 0.19 37.52 *** year90 0.07 1.31 lsize -0.01 -3.01 *** year91 0.02 2.47 ** callable 0.05 4.93 *** year92 0.01 0.96 debt_ipo 0.02 0.90 year93 0.02 2.86 *** public -0.02 -1.81 * year94 0.00 3.21 *** year95 -0.01 3.37 *** year96 0.01 3.03 *** year97 0.00 2.26 ** year98 0.00 0.82 Obs: 3,407 R-Squared 0.43

CHAPTER 2 UNDERWRITING FEES

To test whether the junk bond market from 1977-1989 follows the structure- performance hypothesis, we first develop a theoretical model of a market dominated by a single underwriter. Based on the results from our model, we empirically examine the fees charged to underwrite junk from 1977-1999. We use linear regressions to control for common factors that affect underwriting fees and treasury spreads to isolate pricing effects specific to the underwriter and the time period.

Theoretical Model

Bruck (1989), Stewart (1992) and Stein (1992) document Milken’s dominance in the after-market trading of junk bonds and his superior reputation as an innovator. To see how this would influence underwriting fees, we develop a model where underwriters compete against a single underwriter with a dominant advantage in quality and knowledge of the market. We model Drexel’s quality advantage by giving our single underwriter a permanent edge in delivering payoff to the issuers. In our model, other underwriters can approach the quality of the dominant underwriter but never overcome it.

We model Drexel’s superior knowledge of the market by allowing the single underwriter the right to see the fees of his competitors before announcing his own fee.

Suppose there is a set of underwriters with quality value q, which is public knowledge. q ranges from [0,1]. There is a single dominant underwriter whose quality = 1.

There are also a number of secondary underwriters, whose quality (q) is a function of the cumulative number of deals (δ) they underwrite as follows:

24 25

1 q = 1− (2) 2 +δ

There are 2 kinds of issuers who will choose an underwriter:

High quality issuers (QH): Will only issue through underwriters who are high-

quality. A high-quality underwriter is any underwriter who can provide a payoff > ½. A high-quality issuer will choose the lowest fee among available high-quality underwriters.

Low quality issuers (QL): Choose the lowest fee, regardless of the payoff.

Let NH be the number of high-quality issuers and NL be the number of low quality

issuers. Let market is dominated by high quality needs, so that

N H > 2* N L (3)

Let f be the fee charged by the underwriters, and let ν be the number of issues

 1 underwritten by an underwriter. The underwriters may choose any fee from 0,  . The  2 payoff for the underwriter is the fee times the number of issues done:

π Underwriter = f *ν (4)

The payoff for the issuers is the quality received minus the fee paid to the

underwriter:

π issuers = q − f (5)

The first mover has the following advantages: (1) the top-underwriter (first mover)

earns a permanent quality perception of q=1, which is not dependent on the number of

deals he completes (never changes). (2) The top-underwriter gets to see all other

underwriter’s fees before setting his own.

The game is played as follows:

1. Quality is revealed. 2. The secondary-underwriters choose their respective fees, fs. 3. The top-underwriter responds with fee of ftop. 4. Issuers choose the best quality/fee combination based on their type.

Now consider the case of underwriters competing against an established first mover

who has an unbeatable quality advantage and the right to respond to your fees. For

example, assume that the first game is played with the following conditions:

Underwriter Quality

A 1.00

B 0.50

C 0.50

D 0.50

In choosing their fee, the low-quality underwriters have to consider not only how to

maximize profits but their ability to play in the future. If they are undercut on fees, they

will underwrite fewer issues and be at a quality disadvantage in the future. They also

must consider that the dominant underwriter will respond to their fees and does not have

to worry about earning quality through issues.

Note that the top-underwriter, by virtue of permanent higher quality, can always undercut the secondary underwriters and capture the entire market. This will also allow him to keep the competition from earning quality through issues. However, since the majority of the market is made up of issuers who need high quality, he can choose not to compete for the low quality issues. He will do this as long as he can capture the high- quality market with a higher fee and earn more than if he captured the entire market with a lower fee. That is, when the following condition is satisfied:

f top N H > f s (N H + N L) (6)

meaning

f top N H f s < (7) (N H + N L)

This condition is an upper bound on the fee the secondary underwriters may charge without inviting competition from the top-underwriter. The secondary underwriters also have to worry about being undercut by each other and being at a quality disadvantage in the next game.

In the example game above, the top-underwriter faces no competition for the high- quality market, since even at fs=0 the secondary underwriters cannot provide a payoff >

½. If the secondary-underwriters charge any positive fee, they risk being undercut by their competitors and being at a quality disadvantage in the future. For the top- underwriter, any fee below 0.50 (say 0.50 – α, where α is a very small positive number) will allow him to capture the entire high-quality market. The resulting equilibrium is:

ftop = 0.5-α πtop = (0.5-α) * NH (8)

fs = 0 πs = 0.00* NL = 0 (9)

The low-quality market will be split among secondary-underwriters B, C and D, according to

N L ν = (10) si 3

The secondary underwriters will earn zero, but gain quality for the next game. The next game will be played as follows:

Underwriter Quality

A 1.00

B 1 1− N L 3

C 1 1− N L 3

D 1 1− N L 3

Now the secondary issuers can charge positive fees and still provide quality to the

issuers. If they choose a fee so low that the payoff from issuing through them > 0.50,

however, they will be invite competition from the top-underwriter in a game they cannot

win. The top-underwriter can always undercut them and gain the entire market, with the resulting blow to their ability to earn quality for the future. This puts a lower limit on the fee they can charge without inviting competition from the top-underwriter for the low- quality market. Therefore, the secondary underwriters will choose a fee such that

1 q − f = (11) s s 2

They cannot charge a lower fee because they cannot compete with the top-

underwriter, while they cannot charge a higher fee without risking being undercut by

another secondary underwriter.

If there are n secondary underwriters and the game has been played ω times, then

1 1 f s = − (12) 2 ωN L n

1 Note that as the game continues, ω → ∞ and f → . Of course, at some point fs s 2 becomes large enough that it is worth it for the top-underwriter to lower his fees and capture the entire market (equation 7).

Suppose, however, that there is a penalty for capturing all of the market (anti-trust problems, diseconomies of scale, etc.) Let this penalty be of amount ρ. If this penalty is large enough, the top-underwriter will choose to charge a fee equal to fs and split the market rather than capture it all. This condition is

 N L  ()f s −α ()N H + N L − ρ < f s  N H +  (13)  n 

 N L  ρ > f s  N L −  (14)  n 

If condition (15) holds, the top-underwriter would rather charge f top = f s and split the market rather than charge anything lower and capture it all. In this case, once ω is

1 large enough, then f = f → . top s 2

Next, consider the case where after a number of games, the dominant underwriter is exogenously removed from the market. Now the secondary underwriters face an entirely different situation: they no longer need worry about being undercut by the top-

underwriter, and they can now compete for the NH issues. If an underwriter chooses a

high fee, he can be shut out of the market should the other underwriters charge a lower fee, and be at a quality disadvantage in the next game. The result is that fees are driven to zero.

Our model shows how a single underwriter with a significant edge in quality can charge higher fees while keeping a large market share. Other underwriters will choose to charge fees that do not invite competition from the top-underwriter. As they gain experience in the market and can deliver higher payoffs to issuers, their fees grow to approach those of the top-underwriter. When the top-underwriter is exogenously removed from the market, however, the resulting competitive pressures drive fees much lower.

Empirical Analysis

Our theoretical model suggests that a market dominated by a single underwriter with a distinct quality advantage can maintain high prices even as other underwriters enter the market. Without the dominant player, however, prices will revert to more competitive levels. Using a sample of bonds issued from 1977-1999, we look for evidence of the structure-performance hypothesis by comparing Drexel’s underwriting fees to other underwriters from 1977-1989, and comparing the fees from the Drexel period to fees charged after Drexel’s demise in 1990.

Figure 2-1 is a graph of the median underwriter fee (as a percent of principal) in the junk and investment market from 1977-2002. The median fee fell dramatically in the junk

market after Drexel’s bankruptcy in 1990, from 3.50 percent in 1989 to 2.5 percent in

1992. For investment grade issues, however, the decline in underwriter fees began earlier,

falling from 0.9 percent in 1979 to 0.7 percent in 1981. After that, the median fee was 0.7

percent for every year between 1981 and 1997, except for 1983 when it was 0.8 percent.

Data on market concentration for underwriting non-convertible debt during our

sample period is presented in Table 2-1 for investment-grade issues and Table 2-2 for

junk. Market concentration is as the amount of principal issued by a specific underwriter

that year divided by the total amount of principal issued that year. The top underwriter is

the underwriter with the highest percentage of principal issued in that year. We also

present the concentration of the top two underwriters.

The concentration of the junk market in the 1980s is evident in Table 2-2. From

1980-1989, the top junk underwriter share of principal issued averaged 47 percent and the top two underwriters averaged 61 percent. By contrast, the 1980s investment grade market concentrations averaged 25 percent for the top underwriter and 43 percent for the top two underwriters. From 1990-1999, the averages are virtually the same for the investment market (25 percent for the top underwriter and 41 percent for the top two), while the concentration in the junk market fell dramatically to an average of 25 percent for the top underwriter and 39 percent for the top two. This pattern is seen in the

Herfindahl index, also reported in Table 2-1 and Table 2-8. We calculate the Herfindahl index as follows:

n 2 Herf year = ∑ MarketConcentrationi (16) 1

where market concentration is in percent. For investment-grade issues, the average

Herfindahl index from 1980-1989 is 1,544. This is virtually unchanged from the 1990-

1999 average of 1,467. For junk issues however, the average Herfindahl index from

1980-1989 is 2,851. This falls to an average of 1,343 for 1990-1999.

Information specific to the junk bond market and the dominance of Drexel is shown

in Table 2-3. The results are reported for the junk issues of Drexel versus the issues from

all other underwriters. Between 1977 and 1989, Drexel issued 43 percent of the junk

bonds and 45 percent of the junk principal in our sample. In 1981 and from 1984-1986,

Drexel issued more principal in our sample than all other underwriters combined.

The contrast between the concentration of the 1980s junk market with the 1990s

junk market is evident in Table 2-4, which shows the percent of total principal issued by

each of the top ten underwriters in each decade. In the 1980s, Drexel issued 45 percent of

the total principal in our sample; First Boston, the number two underwriter, issued only

10 percent. For the 1990s, however, the share of principal issued by the top underwriter,

Donaldson Lufkin & Jenrette, was only to 18 percent, with the number two underwriter,

Merrill Lynch, close behind at 13%.

Interestingly, as other underwriters began to gain market share later in the 1980s,

their fees drift toward Drexel’s. Drexel’s median fee is 3.5 percent in 4 of the 6 years

following 1983, the exceptions still being close to 3.5 percent (3.3 percent in 1985 and

3.4 percent in 1986). The non-Drexel median underwriting fee started below this level, at

2.5 percent in 1983, but climbed to match Drexel’s fee of 3.5 percent by 1987, and

remained at that level in 1988 and 1989. In 1984, 24 percent of junk bonds in our sample

paid an underwriter fee of exactly 3.50 percent. This rose to 36 percent in 1987, and by

1989 49 percent of the junk bonds in our sample paid an underwriting fee of exactly 3.50

percent. This pronounced effect is visible in Figure 2-2, which graphs the median fee of

Drexel and non-Drexel bonds during our sample period, and shows how the median non-

Drexel fee increased after 1983 until it was equal to the median Drexel fee from 1987-

1989.

Other changes during the 1980s do not account for the increase in median fees.

Between 1983 and 1989, median issue size increased from $60 million in 1983 to $169

million in 1989. Maturities grew shorter, starting at a median of 15 years to maturity in

1983 and declining to a median maturity of 10 years in 1989. Previous research has

documented economies of scale in the underwriting process during this period (see Lee,

Lochhead, Ritter and Zhao 1996), so merely increasing the size of issues should not result

in higher fees. Shorter maturities tend to lower the risk of the offering, and ceteris

paribus, should decrease underwriting fees.

The statistics show that Drexel dominated the market for newly issued junk bonds in the 1980s, although their dominance was less in 1989 than it was in 1984. We have also shown that median fees in the 1980s junk market were significantly higher than fees in the 1990s. The upward trend of fees in the second half of 1990 toward a median 3.5 percent, matching the median fee charged by Drexel for most of the decade, follows the pattern suggested by our theoretical model: as other underwriters gained experience in the junk market, they were able to charge fees closer and closer to Drexel’s.

Based on these summary statistics, we have reason to believe that the junk bond market in the 1980s was not competitive due to the dominance of a single underwriter and the ability of other underwriters to match Drexel’s fees. This non-competitive environment was reversed in the 1990s, as the market became less concentrated following the bankruptcy of Drexel. We test this hypothesis in the next section.

Regression Analysis

Underwriting Fees

We test whether Drexel’s high market concentration translated into the power to

charge higher fees by running a multivariable regression to control for other influences.

Factors such as size, credit rating, and maturity have been shown to impact underwriter fees for newly issued junk bonds. We regress the underwriting fee as a percent of principal on a set of control variables as follows:

UnderwriterSpread = α 0 + β cr CreditDummies + β split SplitDummiess + β 1 TopUW + (17) β 2 Log(Maturity) + β 3 Log(Size) + β 4 Callable + β 5 TopUWIPO + β 6 NonTopIPO +

β 7 Public + β ind IndustryDummies + β year YearDummies

where the control variables are defined as follows:

CREDIT: A dummy variable for each credit rating. For example, DummyB

equals 1 if the issue is rated B, and equals 0 otherwise. I omit the

AAA dummy in the investment grade sample, so the credit

dummy coefficients are relative to a AAA bond.

SPLIT: A dummy variable indicating that Moody’s rating of the issue is

not the same as the S&P rating. I omit the Baa/Ba split dummy

in the junk bond sample, so the credit dummy coefficients are

relative to a Baa/Ba split rated bond.

TopUW: A dummy variable that equals 1 if the lead underwriter issued

more principal in the junk bond market than any other

underwriter in that year; equals 0 otherwise.

LMAT: The natural log of the years to maturity of the issue

LSIZE: The natural log of the principal amount of the issue

CALLABLE: A dummy variable that equals 1 if the issue is callable, and

equals 0 otherwise.

TopUWIPO A dummy variable that equals 1 if the issue is the first debt issue

by the issuer since 1970 and is issued by the top-uw; equals 0

otherwise.

NonTopUWIPO A dummy variable that equals 1 if the issue is the first debt issue

by the issuer since 1970 and is issued by a nontop-uw; equals 0

otherwise.

Public A dummy variable equal to 1 if the issuer was listed on an

exchange in the year of the debt issue; equals 0 otherwise.

INDUSTRY: A dummy variable for each single-digit SIC code. For example,

SIC3 equals 1 if the issuer has a single-digit SIC code of 3,

equals 0 otherwise.

YEAR: A dummy variable for each year except one. For example,

Dummy80 equals 1 if the bond was issued between January 1,

1980 and December 31, 1980, and equals 0 otherwise.

Our hypothesis is that Drexel was able to maintain a high level of underwriting fees due its position as the dominant underwriter, and that fees in the post-Drexel junk bond market were lower. To measure the effect of Drexel in the 1980s junk market, we run this regression for junk bonds issued from 1977-1989 and compare the results to junk bonds issued from 1990-1999. The results are reported in Table 2-5. The coefficients for each

37 decade are fairly similar, except for the coefficient on TOPUW and TopUWIPO. In the

1977-1989 regression, TOPUW is positive and significant (0.25, t=4.12), indicating that

Drexel’s fees were higher by 25 basis points. This type of market power disappeared in the 1990s, however, as the coefficient on TOPUW is small and insignificant.

Interestingly, all categories of underwriters charge extra fees (from 13-20 basis points) for issuing debt-ipos except for Drexel. TopUWIPO is positive and significant only in the 1990-1999 sample, and NontopIPO is positive and significant in both samples. We see this as further evidence of Drexel’s superior knowledge of the junk- bond market. With his legendary work in researching and following both existing and potential junk issuers, Milken may not have faced the information asymmetry problems other underwriters did with debt-ipos.

To test our claim that other underwriters began to match Drexel’s fees in the late

1980s, we run the following regression:

UnderwriterSpread = α 0 + β cr CreditDummies + β split SplitDummiess + β 2 NonTopUW

+ β 3 Log(Maturity) + β 4 Log(Size) +β 4 DebtIPO + β 5 Callable + β 6 Putable + (18)

β ind IndustryDummies + β NTYearr NonTopYear + β year YearDummies

where we introduce new control variables as follows:

NonTopUW: Dummy variable equal to 1 if the issuer is not the top

underwriter in that year; equal to 0 otherwise.

NonTopYear: The interaction between NonTop and Year dummies. For

example, NonTopYear83 equals 1 if the bond was issued in

1983 and issued by a NonTop underwriter; equals 0 otherwise.

The regression is designed so that the difference between the level of fees of top

underwriters and non-top underwriters for each year will be the sum of the NonTopUW

and NonTopYear coefficients. To see why this is so, note that for top underwriters,

equation 2 becomes:

TopUnderwriterSpread = + β CreditDummies + β SplitDummiess + β Log(Maturity) + α 0 cr split 3 (19) β 4 Log(Size)+β 4 DebtIPO + β 5 Callable + β 6 Putable + β ind IndustryDummies + β year YearDummies

since the NonTopUW and NonTopYear dummies are zero. For NonTop underwriters, all of the terms from (18) remain. So to compare two issues that are alike in every way except that one is issued by a top underwriter and the other is not, we subtract

(19) from (18):

NonTopUnderwriterSpread −TopUnderwriterSpread = (20) β 2 NonTopUW + β NTYear NonTopYear

The results, reported in Table 2-6, confirm our assertion that fees charged by other

underwriters grew closer to Drexel’s fees over time. Fees prior to 1984 are different only

in 1980, but after 1983, the first year of a decade of tremendous growth in the junk bond market, Drexel’s fees are higher in every year except 1988. The value of NonTopUW +

NonTopYear is negative and significant from 1984-1987. In 1986, however, the difference has narrowed to a little more than half of what it was in 1984. By 1988 the difference is not significant. As the market began to expand in 1983, Drexel’s fees were initially higher than those of other underwriters, but the difference grew smaller over time and virtually disappeared by 1988.

Evidence from our regressions confirms that underwriter fees were higher in the

1980s then the 1990s. Drexel was able to charge fees higher than other underwriters during the 1980s, but the dominant underwriters in the 1990s were not able to duplicate

Drexel’s pricing power (see Table 2-5). As they gained experience, other underwriters began to match Drexel’s fees in the late 1980s (see Table 2-6). After the collapse of the junk bond market in 1990, however, the market for underwriting junk bonds became much less concentrated and not dominated by a single underwriter (see Table 2-4). As a result, fees fell significantly in 1992 and have remained lower throughout the 1990s.

Treasury Spreads

Our hypothesis is that Drexel was able to maintain higher fees because of their inherent quality advantage. We look for the source of this advantage by drawing upon the results of Livingston, Pratt and Mann (1995). In their sample of 2,725 bonds, including

404 junk bonds, and document that treasury spreads for Drexel issues were lower than non-Drexel bonds.

The fees paid to the underwriter are just one component of the cost of issuing debt.

The yield require to sell the bonds at a given price is the major cost to the issuer. If an underwriter is able to sell your bonds at a lower yield, that is worth a substantial amount in additional fees to the underwriter. For instance, suppose you have to chose between two underwriters, A and B with different fee/yield options. Underwriter A can sell the bonds for par with a yield of 8 percent, underwriter B can sell the bonds at par with a yield of 7.75 percent. If we assume a $10 million issue with 10 years to maturity and a cost of capital of 10 percent, the savings of 25 basis points in yield has a present value of just over $150,000. So the firm would be willing to pay an additional 150 basis points in fees to underwriter B to save the 25 basis points in yield.

To test the power of the top-underwriter in determining treasury spreads in the

1980s vs. the 1990s, we substitute Treasury spread of the issue (in basis points) for the underwriting spread in regressions (16) and (17) again on our sample of junk bonds. The

results from regression (16) are presented in Table 2-7 and regression (17) in Table 2-8.

We do not see a significant ability for Drexel to sell bonds at lower spreads than other underwriters in the 1977-1989 sample. The TopUW coefficient in Table 2-7 is a negative

16.50 basis points, but is not significant (t=1.60). Curiously, the top-underwriter in the

1990-1999 sample has treasury spreads almost 30 basis points higher than other underwriters. The results may suggest that the hardest to sell issues seek out the top- underwriter, and that Drexel was able to sell those better than top-underwriters in the

1990s. Since our regressions control for factors commonly associated with the inherent risk of bonds (credit ratings, issue size, and time to maturity), it is unclear what other factors could produce this result.

The year by year differences in treasury spreads between Drexel and all other underwriters shown in Table 2-8 show marginal support for Drexel’s ability to sell at higher prices. The difference in spreads is large and significant in 1981 (108 basis points), 1983 (63 basis points) and 1986 (29 basis points). The differences in all other years from 1977-1989 are not different from zero. Although these results do not make a strong case for Drexel’s ability to get higher prices for junk bonds issues, we do find that

Drexel’s treasury spreads are lower in 3 of the years and are never higher than other underwriters.

Table 2-1. Market share of investment-grade underwriters 1977-1999. Sample includes all fixed coupon, non-financial, non-regulated bonds issued 1977- 1999 and rated Baa or higher with yield data available from the SDC database. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. If the bond is not rated by either agency it is not included in the sample. Top Underwriter is the underwriter who issued the most principal that year. Shares are percent of principal issued. The Herf index is calculated as the sum of all market shares (in percent) squared.

Percent Share Percent Share of Top of Top 2 Top Underwriter Underwriter Underwriters Herf Index 1977 Morgan Stanley 20 39 1,194 1978 Goldman Sachs 28 43 1,465 1979 Salomon Brothers 28 53 1,852 1980 Morgan Stanley 18 33 1,105 1981 Morgan Stanley 31 52 1,842 1982 Goldman Sachs 26 43 1,447 1983 Goldman Sachs 21 39 1,299 1984 Salomon Brothers 45 58 2,516 1985 Salomon Brothers 34 55 1,929 1986 Goldman Sachs 21 39 1,346 1987 Salomon Brothers 18 36 1,399 1988 Morgan Stanley 17 32 1,222 1989 Salomon Brothers 23 41 1,452 1990 Goldman Sachs 25 42 1,607 1991 Goldman Sachs 26 43 1,515 1992 Goldman Sachs 27 43 1,560 1993 Goldman Sachs 26 41 1,423 1994 Goldman Sachs 27 42 1,592 1995 Goldman Sachs 26 46 1,486 1996 Goldman Sachs 34 47 1,714 1997 Goldman Sachs 19 36 1,200 1998 Morgan Stanley Dean Witter 18 36 1,289 1999 Goldman Sachs 21 36 1,279

Table 2-2. Market share of junk underwriters 1977-1999. Sample includes all fixed coupon, non-financial, non-regulated bonds issued 1977- 1999 and rated Ba or lower with yield data available from the SDC database. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. If the bond is not rated by either agency it is not included in the sample. Top Underwriter is the underwriter who issued the most principal that year. Shares are percent of principal issued. The Herf index is calculated as the sum of all market shares (in percent) squared.

Percent Share Percent Share of Top of Top 2 Top Underwriter Underwriter Underwriters Herf 1977 Drexel, Burnham, Lambert 35 64 2,468 1978 Drexel, Burnham, Lambert 41 53 2,099 1979 Drexel, Burnham, Lambert 34 46 1,763 1980 Drexel, Burnham, Lambert 42 56 2,309 1981 Drexel, Burnham, Lambert 62 62 4,131 1982 Merrill Lynch 57 84 4,102 1983 Drexel, Burnham, Lambert 41 63 2,444 1984 Drexel, Burnham, Lambert 70 76 4,970 1985 Drexel, Burnham, Lambert 57 70 3,563 1986 Drexel, Burnham, Lambert 53 60 3,032 1987 Drexel, Burnham, Lambert 39 55 2,180 1988 Drexel, Burnham, Lambert 41 55 2,197 1989 Drexel, Burnham, Lambert 34 49 1,808 1990 Goldman Sachs 38 38 2,426 1991 Merrill Lynch 44 61 2,439 1992 Goldman Sachs 19 30 1,006 1993 Merrill Lynch 22 37 1,087 1994 Donaldson Lufkin & Jenrette 21 35 1,028 1995 Donaldson Lufkin & Jenrette 21 37 1,044 1996 Donaldson Lufkin & Jenrette 17 32 821 1997 Donaldson Lufkin & Jenrette 18 35 894 1998 Donaldson Lufkin & Jenrette 20 31 973 1999 Donaldson Lufkin & Jenrette 35 50 1,712

Table 2-3. Summary statistics for the junk bond market from 1977-1989. The sample includes all non-convertible, non-financial, non-regulated fixed-coupon bonds rated as junk with yield data available from the SDC database. The top panel reports bonds whose lead underwriter was Drexel Burnham Lambert; the bottom panel reports bonds whose lead underwriter was not Drexel Burnham Lambert. Total Principal Issued, Mean Size and Total Underwriting Fees are in millions of dollars. Mean Maturity is in years. Median and Mean UW Fee is the total underwriting fees for the issue in percent of total principal issued. IPO indicates the issue was the first debt issue by the company since 1970. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Ratings are assigned numerical values as follows: Baa = 4, Ba = 5, B = 6, Caa = 7. Split rated bonds are assigned the average of the two ratings. If the bond is not rated by either agency it is not included in the sample. Drexel Issues Total Median Total Number of Principal Median Median Median Number of Underwriter Underwriting Issues Issued Size Maturity Rating IPOs Fee Fees 1977 5 105 20 15 5.5 4 3.0 3 1978 11 325 20 18 6.0 8 3.3 10 1979 8 238 23 19 6.0 4 3.3 7 1980 7 285 30 17 6.0 6 3.3 9 1981 8 480 63 20 6.0 3 2.7 12 1982 8 205 20 7 6.0 2 2.7 6 1983 13 770 60 15 6.0 9 3.0 24 1984 19 1,825 80 10 6.0 10 3.5 61 1985 31 3,358 70 12 6.0 11 3.3 108 1986 47 9,075 90 10 6.0 22 3.4 310 1987 30 5,863 100 10 6.0 11 3.5 203 1988 31 4,883 105 10 6.0 14 3.5 164 1989 17 4,016 151 10 6.0 10 3.5 138 235 31,426 114 1,054

Non-Drexel Issues Total Median Total Number of Principal Median Median Median Number of Underwriter Underwriting Issues Issued Size Maturity Rating IPOs Fee Fees 1977 6 193 30 18 5.5 2 2.4 4 1978 19 460 25 20 6.0 14 3.0 14 1979 15 452 25 20 6.0 9 3.0 13 1980 11 390 30 20 6.0 4 3.0 11 1981 4 290 80 20 5.5 3 2.4 7 1982 8 556 58 20 6.0 4 2.9 16 1983 12 1,105 83 20 5.3 3 2.5 27 1984 15 785 35 15 6.0 8 3.0 21 1985 32 2,490 50 10 6.0 20 2.8 66 1986 66 8,051 100 10 6.0 41 3.3 233 1987 50 8,998 125 10 6.0 28 3.5 289 1988 34 6,930 155 10 6.0 19 3.5 234 1989 38 7,869 173 10 6.0 16 3.5 257 310 38,570 171 1,190

Table 2-4. Percent of total junk principal issued by the top 10 underwriters. The sample consists of all unregulated, non-financial, fixed-coupon bonds issued from 1980-1999 rated as junk by Moodys or Standard and Poor’s with yield data available in SDC. Panel A reports junk issues from 1980-1989, Panel B reports issues from 1990-1999. 1977-1990 1990-1999 Percent Percent of of principal principal Underwriter issued Underwriter issued Drexel, Burnham, Lambert 45 Donaldson Lufkin & Jenrette 18 First Boston Corp 10 Merrill Lynch & Co 13 Morgan Stanley 8 Goldman Sachs & Co 8 Goldman Sachs & Co 7 Lehman Brothers 6 Merrill Lynch Capital Markets 7 Chase Securities 6 Salomon Brothers 6 Morgan Stanley & Co 5 Bear Stearns & Co 3 Salomon Brothers 5 Kidder Peabody & Co 2 Bear Stearns & Co 4 Shearson Lehman Brothers 2 BT Securities Corp 4 Merrill Lynch & Co Inc 2 CS First Boston Corp 4

Table 2-5. OLS regression of underwriting spreads for junk bonds. The sample consists of all bonds issued between January 1, 1977 and December 31, 1999 with yield data available from the SDC database and rated Ba or lower by Moody’s. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Bonds not rated by either agency are not included in the sample. The regression takes the following form: UnderwriterSpread = + β CreditDummies + β SplitDummiess + β TopUW + β Log(Maturity) + β Log(Size) α 0 cr split 1 2 3 + β 4 Callable + β 5TopUWIPO + β 6 NonTopIPO + β 7 Public + β ind IndustryDummies + β year YearDummies Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued. Credit Dummies are dummy variables for each Moody credit level. Splits are dummy variables indicating the bond was rated differently by each agency. For the 1977- 1989 sample, TopUW is a variable that equals 1 if the lead underwriter of the issue was Drexel Burnham Lambert, and equals 0 otherwise. For the 1990-1999 sample, TopUW is a variable that equals 1 if the lead underwriter of the issue issued more principal in the junk bond market than any other underwriter in that year; it equals 0 otherwise. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. TopIPO and NonTopIPO are dummy variables that equal 1 if the lead underwriter of the issue was a Top (NonTop) underwriter in the year of the issue, and equals 0 otherwise. Public is a dummy variable that equals 1 if the issuer’s equity was listed on an exchange in the year of the issue, and equals 0 otherwise. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. Industry Dummies are dummy variables (not reported) for each one-digit SIC code. Year Dummies (not reported) are dummy variables for each year from 1977-1999. *, **, *** indicate significance at the 10%, 5% and 1% level respectively. The omitted dummies are SplitBaa/Ba, year89 and year99.

1977 - 1989 1990-1999 Parameter Estimate t-value Estimate t-value Intercept 2.11 8.35 *** 0.56 2.79 *** dummyC 1.76 9.04 *** 2.00 9.19 *** dummyB 1.43 10.88 *** 1.70 25.03 *** dummyBa 0.74 5.21 *** 1.02 14.76 *** split5_6 1.05 7.27 *** 1.52 18.59 *** split6_7 1.49 10.08 *** 1.86 14.43 *** lmat -0.04 -0.59 0.16 2.53 ** lsize -0.10 -3.47 *** -0.07 -2.49 ** callable 0.41 3.25 *** 0.22 4.57 *** top_uw 0.25 4.12 *** -0.01 -0.17 nontop_ipo 0.13 2.26 ** 0.19 5.05 *** topuw_ipo 0.06 0.97 0.20 2.40 ** public -0.05 -1.05 -0.04 -1.10 Obs: 545 967 R-Squared 0.49 0.61

Table 2-6. OLS regression to determine difference between underwriter spread of the top underwriter and those of all other underwriters 1977-1989. The sample consists of all bonds issued between January 1, 1977 and December 31, 1989 and rated Ba or lower by Moody’s. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Bonds not rated by either agency are not included in the sample. The regression takes the following form:

UnderwriterSpread = α 0 + β cr CreditDummies + β split SplitDummiess + β 2 NonTopUW + β 3 Log(Maturity) + β 4 Log(Size) +β 4 DebtIPO

+ β 5 Callable + β 6 Putable + β ind IndustryDummies + β NTYearr NonTopYear + β year YearDummies 1977-1989 Parameter Estimate t-value Parameter Estimate t-value Parameter Estimate p-value Intercept 2.63 9.61 *** dummyC 1.72 8.94 *** dummyB 1.41 10.89 *** dummyBa 0.68 4.84 *** split5_6 1.02 7.12 *** split6_7 1.46 9.96 *** lmat -0.07 -1.11

lsize -0.13 -4.37 *** 46 callable 0.42 3.38 *** nontop -0.27 -1.99 ** debt_ipo 0.09 1.99 ** public -0.07 -1.35 nondrex77 -0.01 -0.04 year77 -0.69 -2.77 *** NonTop+NonTopyear77 -0.28 0.31 nondrex78 0.48 2.16 ** year78 -0.61 -3.17 *** NonTop+NonTopyear78 0.21 0.23 nondrex79 0.30 1.22 year79 -0.48 -2.25 ** NonTop+NonTopyear79 0.03 0.89 nondrex80 0.04 0.17 year80 -0.51 -2.34 ** NonTop+NonTopyear80 -0.23 0.31 nondrex81 0.46 1.47 year81 -0.95 -4.56 *** NonTop+NonTopyear81 0.19 0.50 nondrex82 0.86 3.07 *** year82 -1.26 -5.94 *** NonTop+NonTopyear82 0.59 0.02 ** nondrex83 0.27 1.15 year83 -0.64 -3.54 *** NonTop+NonTopyear83 0.00 1.00 nondrex84 -0.30 -1.44 year84 -0.15 -0.94 NonTop+NonTopyear84 -0.57 0.00 *** nondrex85 -0.21 -1.17 year85 -0.24 -1.62 NonTop+NonTopyear85 -0.48 0.00 *** nondrex86 -0.06 -0.37 year86 -0.10 -0.78 NonTop+NonTopyear86 -0.33 0.00 *** nondrex87 -0.05 -0.27 year87 0.02 0.13 NonTop+NonTopyear87 -0.32 0.00 *** nondrex88 0.24 1.34 year88 -0.21 -1.51 NonTop+NonTopyear88 -0.03 0.79 Obs: 545 R-Squared: 0.53

Table 2-6. Continued Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued. Credit Dummies are dummy variables (not reported) for each Moody credit level. Splits are dummy variables (not reported) indicating the bond was rated differently by each agency. NonTopUW is a dummy variable indicating that the lead underwriter of the issue is not the issuer of the largest amount of principal in the market for that year. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. DebtIPO is a dummy variable equals 1 if the issue was the first debt issue by the company since 1970; equals 0 otherwise. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. Putable is a dummy variable equals 1 is the bond issued is putable, equals 0 otherwise. Industry Dummies are dummy variables (not reported) for each one-digit SIC code. NonTopYear is the product of the NonTop dummy variable and the Year dummy. Year Dummies are dummy variable for each year from 1977-1989. For example, dummy77 equals 1 if the bond was issued in 1977; equals 0 otherwise. The value of NonTop+Year equals the distance between the UW fee of the top underwriter and all others. p-values are based on an F-test of NonTop+NonTopYear equals 0. *, **, *** indicate significance at the 10%, 5% and 1% level respectively. 47

Table 2-7. OLS regression of treasury spreads for junk bonds. The sample consists of all bonds issued between January 1, 1977 and December 31, 1999 with yield data available from the SDC database and rated Ba or lower by Moody’s. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Bonds not rated by either agency are not included in the sample. The regression takes the following form: TreasurySpread = + β CreditDummies + β SplitDummiess + β TopUW + β Log(Maturity) + β Log(Size) α 0 cr split 1 2 3 + β 4 Callable + β 5TopUWIPO + β 6 NonTopIPO + β 7 Public + β ind IndustryDummies + β year YearDummies Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued. Credit Dummies are dummy variables for each Moody credit level. Splits are dummy variables indicating the bond was rated differently by each agency. For the 1977- 1989 sample, TopUW is a variable that equals 1 if the lead underwriter of the issue was Drexel Burnham Lambert, and equals 0 otherwise. For the 1990-1999 sample, TopUW is a variable that equals 1 if the lead underwriter of the issue issued more principal in the junk bond market than any other underwriter in that year; it equals 0 otherwise. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. TopIPO and NonTopIPO are dummy variables that equal 1 if the lead underwriter of the issue was a Top (NonTop) underwriter in the year of the issue, and equals 0 otherwise. Public is a dummy variable that equals 1 if the issuer’s equity was listed on an exchange in the year of the issue, and equals 0 otherwise. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. Industry Dummies are dummy variables (not reported) for each one-digit SIC code. Year Dummies (not reported) are dummy variables for each year from 1977-1999. *, **, *** indicate significance at the 10%, 5% and 1% level respectively. The omitted dummies are SplitBaa/Ba, year89 and year99. 1977 - 1989 1990-1999 Parameter Estimate t-value Estimate t-value Intercept 336.05 7.84 *** 417.73 10.15 *** dummyC 213.54 6.39 *** 355.25 8.34 *** dummyB 184.90 8.27 *** 242.75 17.87 *** dummyBa 80.41 3.32 *** 95.88 7.00 *** split5_6 120.57 4.91 *** 144.36 8.91 *** split6_7 205.02 8.14 *** 352.31 13.05 *** top_uw -16.50 -1.60 28.98 2.46 ** nontop_ipo 1.82 0.19 38.76 4.93 *** topuw_ipo 20.14 1.85 * 17.01 1.02 lmat 25.22 2.37 ** -40.25 -3.07 *** lsize -7.95 -1.61 -19.95 -3.65 *** callable -8.45 -0.41 47.65 4.65 *** public -21.02 -2.55 ** -18.19 -2.38 ** Obs: 524 901 R-Squared: 0.61 0.58

Table 2-8. OLS regression to determine difference between treasury spreads of the top underwriter and those of all other underwriters 1977-1989. The sample consists of all bonds issued between January 1, 1977 and December 31, 1989 and rated Ba or lower by Moody’s. Bonds not rated by Moody’s but rated by S&P are assigned the corresponding Moody’s rating. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Bonds not rated by either agency are not included in the sample. The regression takes the following form: TreasurySpread = + β CreditDummies + β SplitDummiess + β NonTopUW + β Log(Maturity) + β Log(Size) +β DebtIPO α 0 cr split 2 3 4 4 + β 5 Callable + β 6 Putable + β ind IndustryDummies + β NTYearr NonTopYear + β year YearDummies 1977-1989 Parameter Estimate t-value Parameter Estimate t-value Parameter Estimate p-value Intercept 358.70 7.66 *** dummyC 223.91 6.64 *** dummyB 189.36 8.40 *** dummyBa 78.52 3.20 *** split5_6 119.70 4.84 *** split6_7 204.08 8.01 *** lmat 20.36 1.85 * lsize -7.88 -1.57 callable -8.32 -0.40 nontop -31.78 -1.36 49 debt_ipo 7.26 0.95 public -21.75 -2.63 *** nondrex77 -7.26 -0.14 year77 -184.73 -4.44 *** NonTop+NonTopyear77 -39.04 0.41 nondrex78 17.62 0.44 year78 -228.33 -6.86 *** NonTop+NonTopyear78 -14.17 0.66 nondrex79 11.15 0.23 year79 -217.92 -5.24 *** NonTop+NonTopyear79 -20.63 0.62 nondrex80 7.90 0.18 year80 -92.38 -2.52 ** NonTop+NonTopyear80 -23.88 0.53 nondrex81 139.92 2.63 *** year81 -129.34 -3.70 *** NonTop+NonTopyear81 108.14 0.02 ** nondrex82 83.37 1.76 * year82 -91.56 -2.56 ** NonTop+NonTopyear82 51.59 0.21 nondrex83 94.65 2.37 ** year83 -245.77 -8.13 *** NonTop+NonTopyear83 62.86 0.05 * nondrex84 51.64 1.44 year84 -230.76 -8.64 *** NonTop+NonTopyear84 19.86 0.47 nondrex85 38.01 1.23 year85 -137.48 -5.59 *** NonTop+NonTopyear85 6.23 0.76 nondrex86 60.65 2.18 ** year86 -24.51 -1.09 NonTop+NonTopyear86 28.86 0.06 * nondrex87 29.44 0.99 year87 -81.61 -3.39 *** NonTop+NonTopyear87 -2.34 0.90 nondrex88 21.49 0.70 year88 -87.47 -3.66 *** NonTop+NonTopyear88 -10.29 0.60 Obs: 524 R-Squared: 0.62 Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued.. Credit Dummies are dummy variables (not reported) for each Moody credit level. Splits are dummy variables (not reported) indicating the bond was rated differently by each agency. NonTopUW is a dummy variable indicating that the lead underwriter of the issue is not the issuer of the

Table 2-8. Continued largest amount of principal in the market for that year. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. DebtIPO is a dummy variable equals 1 if the issue was the first debt issue by the company since 1970; equals 0 otherwise. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. Putable is a dummy variable equals 1 is the bond issued is putable, equals 0 otherwise. Industry Dummies are dummy variables (not reported) for each one-digit SIC code. NonTopYear is the product of the NonTop dummy variable and the Year dummy. Year Dummies are dummy variable for each year from 1977-1989. For example, dummy77 equals 1 if the bond was issued in 1977; equals 0 otherwise. The value of NonTop+Year equals the distance between the UW fee of the top underwriter and all others. p-values are based on an F-test of NonTop+NonTopYear equals 0. *, **, *** indicate significance at the 10%, 5% and 1% level respectively. 50

4.0

3.5

3.0 l a p 2.5 of Princi

t 2.0 cen 1.5 Fee Per 1.0

0.5 Investment-grade Junk

0.0 1977 1982 1987 1992 1997

Year Issued

Figure 2-1. Median underwriter fees as a percent of principal issued.

3.6

3.4

3.2

2.8

2.6

2.4

2.2 NonDrexel Drexel

2 1976 1978 1980 1982 1984 1986 1988 1990

Figure 2-2. Median junk bond underwriter fees as a percent of principal 1977-1989.

CHAPTER 3 COMMERCIAL BANK ENTRY

Gande, Puri and Saunders (1999) (henceforth GPS) also examine underwriting fees for newly issued bonds from 1985-1996. They conclude that fees fell in the 1990s due to the entry of commercial banks into the bond underwriting market through Section 20 subsidiaries, and point out that the decline in fees is mainly in junk bonds. They find no evidence that commercial bank fees are lower than investment bank fees. Instead, they assert that investment banks lowered fees in response to commercial bank entry. They document that underwriting fees are negatively related to both a time dummy indicating the entry of commercial banks and the market share of the commercial banks. We offer evidence that GPS results are actually due to bank entry occurring at the same time as

Drexel’s exit from the market for new issue junk.

The Bank Act of 1933 (commonly know as Glass-Steagall) was a response to the chaos in the US financial system that resulted from the Great Depression. It was widely believed at the time that ownership and underwriting of securities by banks created an inherent conflict of interest and encouraged risky and sometimes unethical actions on the part of banks. The law was intended to separate the activities of commercial banking

(taking deposits and making loans) from those of investment banks (issuing and trading in securities).

Ironically, just two years after he led the effort to enact the law, Carter Glass helped lead an effort to have it repealed, saying "he thought it was a mistake and an overreaction." His effort failed, and it was over 60 years before the law was finally

53 54

repealed in 1999. By the time of the official repeal, however, the Federal Reserve’s re-

interpretation of Section 20 of the act had already allowed commercial banks to

underwrite corporate debt.

Section 20 of the Bank Act of 1933 forbids member banks from affiliating with a

company “engaged principally” in the “issue, flotation, underwriting, public sale, or

distribution at wholesale or retail or through syndicate participation of stocks, bonds,

debentures, notes, or other securities.” In 1987, the Federal Reserve ruled that “engaged

principally” did not strictly prohibit banks from affiliating with companies involved in

investment banking activities. Banks were allowed to have Section 20 subsidiaries which

could underwrite municipal revenue bonds and commercial paper, as long as the revenues from those subsidiaries did not exceed 5% of the banks total revenues. In 1989, the Fed raised the revenue limit to 10% and added corporate bond underwriting to the activities permitted by section 20 subsidiaries. In 1991, JP Morgan became the first commercial bank since the passage of Glass-Steagall to underwrite a corporate bond issue.

A number of studies have examined the role of commercial banks in the market for underwriting securities. Puri (1994) examines the default rate of securities issued prior to the passage of Glass-Steagall, and finds that securities underwritten by commercial banks had a lower default rate than those underwritten by investment banks. Gande et al. (1997) find that banks provide a certification effect and obtain lower yields for their debt issues.

Puri (1999) presents a model that demonstrates how commercial banks can better certify an issue than investment banks.

To determine the impact of commercial bank entry on junk bond fees we compare the bonds underwritten by commercial banks to those underwritten by investment banks.

Summary statistics for the bonds in our sample are shown in Table 3-1. An issue by

commercial banks tends to be smaller in size, but the size of their issues gets closer to

that of investment banks later in the decade. The median fee charged by commercial

banks is equal to or higher than the corresponding investment bank fee for every year

from 1991-1999 except for 1993 and 1999.

The market share of commercial banks is also presented in Table 3-2. In 1991, the first year a commercial bank issue enters our sample, the commercial banks issued 8 percent of the total junk principal. Their initial share increased to 11 percent in 1992, and stayed about the same until 1995, when commercial banks share of the junk market jumped to 17 percent. This was followed by another jump in 1996 to 33 percent.

In Table 3-3 we report summary statistics for each commercial bank underwriter in our sample. The majority of the principal issued comes from three commercial banks:

Bankers Trust, Chase, and JP Morgan. Chase has 7 percent of the market, followed by JP

Morgan and Banker’s Trust with 4 percent each. Nationsbank and Citicorp each have 2 percent of the market; no other commercial bank has more than 1 percent of the market.

The market share of the top six commercial banks for each year is shown in Table . Prior to 1995, no single commercial bank ever had a market share above 5 percent for the year.

From 1995 to 1999, commercial banks as a whole made big gains in market share, but no single bank ever maintained a large share. For example, Chase’s market share is 18 percent in 1997, but then falls to 12 percent in 1998 and 9 percent in 1999.

Based on the summary statistics, we find little to support the argument that commercial bank entry caused lower fees in the 1990s. Commercial bank median fees are the no lower than those charged by investment banks, and for the first half of the decade

commercial banks have a small market share. To be sure that we are properly controlling

for issue characteristics, we replace TOPUW in regression (1) with a dummy variable CB

that equals 1 if the lead underwriter is a commercial bank, and equals 0 otherwise. We

run the regression on the sample of junk bonds issued from 1991-1999. The results are

listed in Table 3-4. The coefficient on the CB dummy is not different from zero,

indicating that after controlling for characteristics specific to each issue, commercial bank

fees were no different than fees charged by other underwriters during that time.

Using a sample of bonds issued from 1985-1996, Gande, Puri and Saunders (1999)

conclude that commercial banks entering the market through section 20 subsidiaries

resulted in lower underwriting fees. By choosing the period 1985-1996, however, GPS

divide the sample roughly in half between the Drexel dominated market (1985-1989) and

the post Drexel market with commercial bank entry (1990-1996). We have shown in

Table that there was a sudden sharp decline in fees for underwriting junk in 1991, and

that the level of fees remained virtually the same from 1991-1996. This coincides

perfectly with GPS claim that bank entry in 1989 resulted in lower fees. However, it also

coincides perfectly with Drexel’s exit from the market.2

GPS also show that underwriting fees are negatively related to the market share of commercial banks. They do this by including the variable LN_MKTSHARE in their regressions, the natural log of the market share of the commercial banks the year the bond was issued. Since this variable is zero for the period from 1985-1990, and positive every year after, it is also correlated with the presence of Drexel.

2 Remember that the 1990 junk bond market was very small, with only 4 bonds in our sample.

It is our claim that fees are lower in the 1990s because of increased competition

following Drexel’s exit. We follow the method of GPS and run the following regression

on our sample of corporate bonds from 1991-1999.

UnderwriterSpread = + β CreditDummies + β Log(Maturity) + β Log(Size)+β DebtIPO + α 0 cr 1 2 3 (1) + β 6 LN (MKTSHARE ) + β ind IndustryDummies + Time

By starting the sample in 1991 instead of 1985, we eliminate the effect of Drexel

exiting the market. We extend the GPS sample from 1996 to 1999 to include issues up to

the repeal of Glass-Steagall in 1999. The results are shown in Table 3-5. The coefficient

on LN_MKTSHARE in this regression is small and insignificant, meaning that without

the higher fees on bonds from the 1985-1989 period in the sample, there is no evidence

that higher bank market share is correlated with lower underwriting fees in the 1990s

junk market.

Based on these statistics, it is difficult to argue that commercial banks had a

dramatic impact on underwriting fees in the bond market. Median fees for underwriting

investment-grade bonds are virtually identical between 1983 and 1997. Fees for issuing

junk are drastically lower after Drexel’s bankruptcy in 1990. However, we find no

evidence that this is due to commercial banks entering the underwriting business. We

document that commercial bank’s underwriting fees were not different from those

charged by other underwriters during the 1990s. Also, commercial banks started with a

small portion of the market and had to deal with regulation that limited the revenues they

could earn from underwriting activities, and therefore limited the amount of market share

they could capture.3

3 From 1989-1995, The Federal Reserve required that no more the 10 percent of a Section 20’s revenue could be from underwriting activities. This limit was raised to 25 percent in 1996.

Table 3-1. Summary statistics: Investment vs. commercial banks. The sample consists of all unregulated, non-financial, fixed-coupon bonds issued from 1990-1999 rated as junk by Moodys or Standard and Poor’s with yield data available in SDC. Bonds rated Baa (BBB) by one agency and Ba (BB) by the other are classified as junk bonds. Bonds not rated by either agency are not included in the sample. The top panel lists issues whose lead underwriter was an investment bank; the bottom panel lists issues whose lead underwriter was a commercial bank. Market Share is in percent. Total Principal and Median Size are in millions of dollars. Median underwriter fee is the total underwriting fee for the issue in percent of total principal issued. Rating is a numerical assignment to Moody’s rating: Aaa equals 1 through Caa equals 7. Split-rated bonds are assigned the average value of the two ratings. Investment Bank Issues Total Median Number of Market Principal Median Median Underwriter Median Issues Share Issued Size Maturity Fee Rating Debt IPOs 1991 25 92 6,459 200 10 2.0 5 3 1992 117 89 18,921 135 10 2.4 6 35 1993 147 89 25,782 125 10 2.8 6 70 1994 80 90 13,914 123 10 2.8 6 44 1995 49 83 9,873 150 10 2.5 6 23 1996 66 66 12,030 150 10 2.8 6 26 1997 82 63 16,315 150 10 2.7 6 35 1998 137 80 28,305 155 10 2.6 6 60 1999 46 60 9,462 173 10 2.5 6 12 749 141,059 308

Commercial Bank Issues Total Median Number of Market Principal Median Median Underwriter Median Issues Share Issued Size Maturity Fee Rating Debt IPOs 199148575150 52.751 1992 16 11 2,318 133 8 2.5 6 7 1993 23 11 3,303 130 10 2.6 6 11 1994 11 10 1,512 100 10 3.0 6 8 1995 13 17 2,036 125 10 2.9 6 4 1996 36 34 6,205 125 10 3.0 6 22 1997 55 37 9,605 150 10 3.0 6 36 1998 40 20 7,070 148 10 2.8 6 26 1999 20 40 6,367 200 8 2.5 6 8 218 189 38,990 123

Table 3-2. Commercial bank underwriters 1991-1999. The sample includes all fixed coupon, non-convertible bonds rated junk by Moodys or S&P with yield data available in SDC. Bonds issued by regulated industries (SIC code 4) and financial firms (SIC code 6) are not included. Bonds not rated by either agency are not included in the sample. The bonds are classified as underwritten by a commercial bank if the lead (or co-lead) underwriter is a commercial bank. Principal amount and size is in millions of US dollars. Rating is a numerical assignment to Moody’s rating: Aaa equals 1 through Caa equals 7. Split-rated bonds are assigned the average value of the two ratings. Total Median Number Market Principal Median Median Underwriter Debt Bank Underwriter of Issues Share Issued Size Rating Fee IPOs BANKAMERICA 5 1 1,127 152 6.0 0.3 2 BANC-BOSTON 2 0 205 103 6.0 3.0 2 BANKERS-TRUST 49 4 7,642 115 6.0 3.0 31 CHASE 63 7 12,560 150 6.0 2.9 33 CHEMICAL 6 1 1,056 116 6.0 3.0 4 CIBC 13 1 2,473 200 6.0 3.0 6 CITICORP 24 2 3,266 128 6.0 2.7 15 DEUTSCHE-BANK 2 0 190 95 6.0 2.9 1 FIRST-CHICAGO 1 0 100 100 6.0 2.8 1 FIRST-UNION 3 0 515 115 6.0 2.8 3 G-G-NATWEST 2 0 260 130 6.0 3.4 2 JP MORGAN 26 4 6,595 150 5.8 2.4 9 NATIONSBANK 22 2 3,002 123 6.0 2.8 14 218 22 38,990 123

Table 3-3. Percent share of the major commercial bank junk bond underwriters 1991- 1999. The sample includes all fixed coupon, non-convertible bonds rated as junk by Moodys or S&P. Bonds issued by regulated industries (SIC code 4) and financial firms (SIC code 6) are not included. Bonds without a Moody’s rating are assigned the corresponding S&P rating. Bonds not rated by either agency are not included in the sample. The bonds are classified as underwritten by a commercial bank if the lead (or co-lead) underwriter is a commercial bank.

Commercial Bank 1991 1992 1993 1994 1995 1996 1997 1998 1999

BANKERS-TRUST 4 4 3 5 8 15 4 - -

CHASE - - 2 1 2 8 18 12 9

CIBC - - - - - 3 6 1 -

CITICORP 2 4 5 3 - 1 1 - -

JP MORGAN 1 3 1 1 4 2 4 1 20

NATIONSBANK - - - - - 2 4 4 1

Table 3-4. Regressions with commercial bank (CB) dummy for issues 1991-1999. The sample includes all fixed coupon, non-convertible bonds rated junk by Moodys or S&P with yield data available in SDC. Bonds issued by regulated industries (SIC code 4) and financial firms (SIC code 6) are not included. Bonds not rated by either agency are not included in the sample. The bonds are classified as underwritten by a commercial bank if the lead (or co-lead) underwriter is a commercial bank.The regression takes the following form: UnderwriterSpread = + β CreditDummies + β Log(Maturity) + β Log(Size)+β DebtIPO + α 0 cr 1 2 3 β 4 Callable + β 5 CB + β ind IndustryDummies + β year YearDummies Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued. Credit Dummies are dummy variables for each Moody credit level. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. DebtIPO is a dummy variable equals 1 if the issue was the first debt issue by the company since 1970; equals 0 otherwise. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. CB is a variable equal to 1 if the lead underwriter of the issue was a commercial bank; equals 0 otherwise. Industry Dummies (not reported) are dummy variables for each single digit SIC code. Year Dummies are dummy variable for each year from 1991-1999. *, **, *** indicates significance at the 10%, 5%, and 1% level respectively. Parameter Estimate t-stat Intercept 0.55 2.74 *** dummyBa 1.03 14.79 *** split4_6 0.10 0.29 dummyB 1.72 25.29 *** split5_6 1.53 18.73 *** split6_7 1.88 14.76 *** dummyC 2.02 9.37 *** lmat 0.15 2.39 ** lsize -0.07 -2.55 ** CB -0.02 -0.46 callable 0.22 4.54 *** debt_ipo 0.21 5.84 *** year90 0.36 1.39 year91 0.28 2.44 ** year92 0.06 0.80 year93 0.19 2.57 ** year94 0.24 2.92 *** year95 0.28 3.12 *** year96 0.23 2.87 *** year97 0.15 1.99 ** year98 0.04 0.61 Obs: 970 R-squared 0.61

Table 3-5. Regressions with commercial bank market share 1991-1999. The sample includes all fixed coupon, non-convertible bonds rated junk by Moodys or S&P with yield data available in SDC. Bonds issued by regulated industries (SIC code 4) and financial firms (SIC code 6) are not included. Bonds not rated by either agency are not included in the sample. The bonds are classified as underwritten by a commercial bank if the lead (or co-lead) underwriter is a commercial bank.The regression takes the following form: UnderwriterSpread = + β CreditDummies + β Log(Maturity) + β Log(Size)+β DebtIPO + α 0 cr 1 2 3 β 4 Callable + β 5 LN (MKTSHARE ) + β ind IndustryDummies + β year YearDummies Underwriter Spread is the total underwriting fee as a percent of the total principal amount issued. Credit Dummies are dummy variables for each Moody credit level. Log(Maturity) is the natural log of time to maturity of the bond in years. Log(Size) is the natural log of the principal size of the issue, in millions of dollars. DebtIPO is a dummy variable equals 1 if the issue was the first debt issue by the company since 1970; equals 0 otherwise. Callable is a dummy variable equals 1 is the bond issued is callable, equals 0 otherwise. LN(MKTSHARE) is the natural log of (1+ market share (in percent)) of commercial banks for that year. Industry Dummies (not reported) are dummy variables for each single digit SIC code. Year Dummies are dummy variable for each year from 1991-1999. *, **, *** indicates significance at the 10%, 5%, and 1% level respectively.

Parameter Estimate t-stat Intercept 0.91 3.01 *** dummyBa 1.03 14.79 *** split4_6 0.11 0.29 dummyB 1.72 25.35 *** split5_6 1.53 18.73 *** split6_7 1.88 14.77 *** dummyC 2.02 9.37 *** lmat 0.15 2.42 ** lsize -0.07 -2.55 ** lnmktshare -0.10 -1.41 callable 0.22 4.53 *** debt_ipo 0.21 5.82 *** year91 0.13 0.91 year92 -0.06 -0.55 year93 0.08 0.76 year94 0.11 1.00 year95 0.20 2.04 ** year96 0.21 2.71 *** year97 0.14 1.90 * year98 -0.02 -0.26 Obs: 970 R-squared 0.61

CHAPTER 4 CONCLUSION

More than a decade after the bankruptcy of Drexel, questions remain about its impact on the market for new issue bonds. Economic research has described concentrated markets by the performance-structure hypothesis, where firms use their high market share to keep prices high, or the efficient-performance hypothesis, where market share is the result of leading firms providing superior quality. We find support for the efficient- structure hypothesis in explaining Drexel’s domination of the junk-bond market from

1977-1989. From 1977-1989, Drexel maintained the dominant market share while charging higher underwriting fees. By the time Milken and Drexel issued their first junk- bond in 1977, they already wielded significant market power because of their role as innovator in the secondary market for junk. This advantage in knowledge and share of the after-market resulted in substantial power in the new-issues market, resulting in higher fees for Drexel issues. As the market grew, other underwriters gained experience in the junk-bond market and over time increased their fees to match Drexel’s.

The market for issuing junk in the 1990s, with Drexel out of the picture, was completely different. Fees to underwrite new junk fell dramatically when the junk market rebounded in 1992-1993, and they stayed low through the decade. In contrast to the

Drexel period, the dominant underwriter in the 1990s could not duplicate Drexel’s position as innovator and dominant player in the secondary market. Hence, Drexel’s power to charge higher fees could not be duplicated.

63 64

Unlike previous researchers, we do not see evidence that this decline is due to commercial banks entering the market. The underwriting fees charged by commercial banks are no different than fees charged by investment banks, and the market share of commercial banks prior to 1995 was about 10%. Until 1996, Federal Reserve regulations limited the percent of total revenues commercial banks could earn from underwriting securities to 15 percent. Commercial banks simply did not have the market power in the early 1990s to impact underwriting fees.

LIST OF REFERENCES

Asquith, Paul, David W. Mullins, and Eric D. Wolff. Original Issue High Yield Bonds: Aging Analysis of Defaults, Exchanges and Calls. Journal of Finance, September 1989, Volume 44, pages 923-952.

Berger, Allen N., Rebecca S. Demsetz and Philip E. Strahan. “The Consolidation of the Financial Services Industry: Causes, Consequences, and Implications for the Future.” Journal of Banking and Finance, Volume 23, pages 135-194.

Berger, Allen N., and Timothy H Hannan. “The Price-Concentration Relationship in Banking.” The Review of Economics and Statistics, 1989, Volume 71, pages 291- 299.

Berman, P and J. W. Michaels. “My story—Michael Milken.” Forbes, March 16, 1992, Volume 149, pages 78-93.

Bleakley, Fred R. “Firms Enter a Risky Field.” New York Times, June 28, 1984, page D1.

Blume, Marshall E., Donald B. Keim, and Sandeep Patel. “Returns and Volatility of Low-Grade Bonds 1977-1989.” Journal of Finance, March 1991, Volume 46, pages 49-74.

Bruck, Connie. The Predators Ball: The Junk-Bond Raiders and the Man Who Staked Them. New York: Simon and Schuster, 1989.

Chen, Hsuan-Chi and Jay Ritter. “The Seven Percent Solution.” Journal of Finance, June 2000, Volume 55, No. 3, pages 1105-1132.

Christie, Williams G, and Paul Schultz. “Why Do NASDAQ Market Makers Avoid Odd- Eighth Quotes?” Journal of Finance, December 1994, Volume 49, pages 1813- 1840.

Cornell, Bradford and Kevin Green. “The Investment Performance of Low-Grade Bond Funds.” Journal of Finance, March 1991, Volume 46, pages 29-48.

Datta, Sudip, Mai Iskandar-Datta, and Ajay Patel. “The Pricing of Initial Offers of Corporate Straight Debt.” Journal of Finance, March 1997, Volume 52, pages 379- 396.

Demsetz, Harold. “Industry Structure, Market Rivalry and Public Policy.” Journal of Law and Economics, April 1973, Volume 16, pages 1-9.

65 66

Dutta, Prajit, and Ananth Madhavan. “Competition and Collusion in Dealer Markets.” Journal of Finance, March 1997, Volume 52, pages 245-276.

Fridson, Martin S. “Do High-Yield Bonds Have an Equity Component?” Financial Management, Summer 1994, Volume 23, pages 82-84.

Gande, Amar, Manju Puri, and Anthony Saunders. “Bank Entry, Competition, and the Market for Corporate Securities Underwriting.” Journal of Financial Economics, October 1999, Volume 54, pages 165-195.

Gilbert, Alton R. “Bank Market Structure and Competition: A Survey.” Journal of Money, Credit and Banking. November 1984, Volume 16, pages 617-645.

Hannan, T.H. “Bank Commercial Loan Markets and the Role of Market Strucutre: Evidence from Surveys of Commercial Lending.” Journal of Banking and Finance, February 1991, Volume 15, pages 133-149.

Harris, Lawrence. “Stock Price Clustering and Discreteness.” Review of Financial Studies, 1991, Volume 4, page 389-415.

Jewell, Jeff and Miles Livingston. “Split Ratings, Bond yields and Underwriting Spreads.” Journal of Financial Research, Summer 1998, Volume 21, pages 185- 204.

Kim, Dongcheol, Darius Palia, and Anthony Saunders. “The Long-Run Behavior of Debt and Equity Underwriting Spreads.” Working paper, January 2003.

Lee, Inmoo, Scott Lochhead, Jay Ritter, and Quanshui Zhao. “The Costs of Raising Capital.” Journal of Financial Research, Spring 1996, Volume 19, No. 1, pages 59- 74.

Livingston, Miles, Hugh Pratt, and Charles Mann. “Drexel Burnham Lambert’s Debt Issues.” Journal of Fixed Income, March 1995, pages 58-75.

Nuemark, David and Steven A. Sharpe. “Market Structure and the Nature of Price Rigitity: Evidence from the Market for Consumer Deposits.” The Quarterly Journal of Economics, May 1992, Volume 107, pages 657-680.

Stein, Benjamin J. A License to Steal. New York: Simon and Schuster, 1992.

Stewart, James B. Den of Thieves. New York: Simon and Schuster, 1992.

Tufano, Peter. “Financial Innovation and First Mover Advantages." Journal of Financial Economics, December 1989, Volume 25, pages 213-240.

Winkler, Matthew. “Junk Bond Market Is Seen Showing 38% Default Rate.” Wall Street Journal, January 25, 1990, page C10.

“Yes, Pardon Milken,” Wall Street Journal, December 13, 2000, page A26.

BIOGRAPHICAL SKETCH

Glenn Williams earned a BS in mechanical engineering from the University of

South Carolina in 1990 and an MBA from Rollins College in 1997. He worked for seven years in the electric utility industry prior to beginning work on his PhD. He has 3 sons,

Benjamin, Daniel and Samuel.